netfilter: nf_conntrack: add allocation flag to nf_conntrack_alloc
[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) 2005-2008 by Pablo Neira Ayuso <pablo@netfilter.org>
8  *
9  * Initial connection tracking via netlink development funded and
10  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
11  *
12  * Further development of this code funded by Astaro AG (http://www.astaro.com)
13  *
14  * This software may be used and distributed according to the terms
15  * of the GNU General Public License, incorporated herein by reference.
16  */
17
18 #include <linux/init.h>
19 #include <linux/module.h>
20 #include <linux/kernel.h>
21 #include <linux/types.h>
22 #include <linux/timer.h>
23 #include <linux/skbuff.h>
24 #include <linux/errno.h>
25 #include <linux/netlink.h>
26 #include <linux/spinlock.h>
27 #include <linux/interrupt.h>
28 #include <linux/notifier.h>
29
30 #include <linux/netfilter.h>
31 #include <net/netlink.h>
32 #include <net/netfilter/nf_conntrack.h>
33 #include <net/netfilter/nf_conntrack_core.h>
34 #include <net/netfilter/nf_conntrack_expect.h>
35 #include <net/netfilter/nf_conntrack_helper.h>
36 #include <net/netfilter/nf_conntrack_l3proto.h>
37 #include <net/netfilter/nf_conntrack_l4proto.h>
38 #include <net/netfilter/nf_conntrack_tuple.h>
39 #ifdef CONFIG_NF_NAT_NEEDED
40 #include <net/netfilter/nf_nat_core.h>
41 #include <net/netfilter/nf_nat_protocol.h>
42 #endif
43
44 #include <linux/netfilter/nfnetlink.h>
45 #include <linux/netfilter/nfnetlink_conntrack.h>
46
47 MODULE_LICENSE("GPL");
48
49 static char __initdata version[] = "0.93";
50
51 static inline int
52 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
53                             const struct nf_conntrack_tuple *tuple,
54                             struct nf_conntrack_l4proto *l4proto)
55 {
56         int ret = 0;
57         struct nlattr *nest_parms;
58
59         nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO | NLA_F_NESTED);
60         if (!nest_parms)
61                 goto nla_put_failure;
62         NLA_PUT_U8(skb, CTA_PROTO_NUM, tuple->dst.protonum);
63
64         if (likely(l4proto->tuple_to_nlattr))
65                 ret = l4proto->tuple_to_nlattr(skb, tuple);
66
67         nla_nest_end(skb, nest_parms);
68
69         return ret;
70
71 nla_put_failure:
72         return -1;
73 }
74
75 static inline int
76 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
77                          const struct nf_conntrack_tuple *tuple,
78                          struct nf_conntrack_l3proto *l3proto)
79 {
80         int ret = 0;
81         struct nlattr *nest_parms;
82
83         nest_parms = nla_nest_start(skb, CTA_TUPLE_IP | NLA_F_NESTED);
84         if (!nest_parms)
85                 goto nla_put_failure;
86
87         if (likely(l3proto->tuple_to_nlattr))
88                 ret = l3proto->tuple_to_nlattr(skb, tuple);
89
90         nla_nest_end(skb, nest_parms);
91
92         return ret;
93
94 nla_put_failure:
95         return -1;
96 }
97
98 static int
99 ctnetlink_dump_tuples(struct sk_buff *skb,
100                       const struct nf_conntrack_tuple *tuple)
101 {
102         int ret;
103         struct nf_conntrack_l3proto *l3proto;
104         struct nf_conntrack_l4proto *l4proto;
105
106         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
107         ret = ctnetlink_dump_tuples_ip(skb, tuple, l3proto);
108         nf_ct_l3proto_put(l3proto);
109
110         if (unlikely(ret < 0))
111                 return ret;
112
113         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
114         ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
115         nf_ct_l4proto_put(l4proto);
116
117         return ret;
118 }
119
120 static inline int
121 ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct)
122 {
123         NLA_PUT_BE32(skb, CTA_STATUS, htonl(ct->status));
124         return 0;
125
126 nla_put_failure:
127         return -1;
128 }
129
130 static inline int
131 ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct)
132 {
133         long timeout = (ct->timeout.expires - jiffies) / HZ;
134
135         if (timeout < 0)
136                 timeout = 0;
137
138         NLA_PUT_BE32(skb, CTA_TIMEOUT, htonl(timeout));
139         return 0;
140
141 nla_put_failure:
142         return -1;
143 }
144
145 static inline int
146 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct nf_conn *ct)
147 {
148         struct nf_conntrack_l4proto *l4proto;
149         struct nlattr *nest_proto;
150         int ret;
151
152         l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
153         if (!l4proto->to_nlattr) {
154                 nf_ct_l4proto_put(l4proto);
155                 return 0;
156         }
157
158         nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
159         if (!nest_proto)
160                 goto nla_put_failure;
161
162         ret = l4proto->to_nlattr(skb, nest_proto, ct);
163
164         nf_ct_l4proto_put(l4proto);
165
166         nla_nest_end(skb, nest_proto);
167
168         return ret;
169
170 nla_put_failure:
171         nf_ct_l4proto_put(l4proto);
172         return -1;
173 }
174
175 static inline int
176 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
177 {
178         struct nlattr *nest_helper;
179         const struct nf_conn_help *help = nfct_help(ct);
180         struct nf_conntrack_helper *helper;
181
182         if (!help)
183                 return 0;
184
185         rcu_read_lock();
186         helper = rcu_dereference(help->helper);
187         if (!helper)
188                 goto out;
189
190         nest_helper = nla_nest_start(skb, CTA_HELP | NLA_F_NESTED);
191         if (!nest_helper)
192                 goto nla_put_failure;
193         NLA_PUT_STRING(skb, CTA_HELP_NAME, helper->name);
194
195         if (helper->to_nlattr)
196                 helper->to_nlattr(skb, ct);
197
198         nla_nest_end(skb, nest_helper);
199 out:
200         rcu_read_unlock();
201         return 0;
202
203 nla_put_failure:
204         rcu_read_unlock();
205         return -1;
206 }
207
208 #ifdef CONFIG_NF_CT_ACCT
209 static int
210 ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
211                         enum ip_conntrack_dir dir)
212 {
213         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
214         struct nlattr *nest_count;
215
216         nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
217         if (!nest_count)
218                 goto nla_put_failure;
219
220         NLA_PUT_BE32(skb, CTA_COUNTERS32_PACKETS,
221                      htonl(ct->counters[dir].packets));
222         NLA_PUT_BE32(skb, CTA_COUNTERS32_BYTES,
223                      htonl(ct->counters[dir].bytes));
224
225         nla_nest_end(skb, nest_count);
226
227         return 0;
228
229 nla_put_failure:
230         return -1;
231 }
232 #else
233 #define ctnetlink_dump_counters(a, b, c) (0)
234 #endif
235
236 #ifdef CONFIG_NF_CONNTRACK_MARK
237 static inline int
238 ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
239 {
240         NLA_PUT_BE32(skb, CTA_MARK, htonl(ct->mark));
241         return 0;
242
243 nla_put_failure:
244         return -1;
245 }
246 #else
247 #define ctnetlink_dump_mark(a, b) (0)
248 #endif
249
250 #ifdef CONFIG_NF_CONNTRACK_SECMARK
251 static inline int
252 ctnetlink_dump_secmark(struct sk_buff *skb, const struct nf_conn *ct)
253 {
254         NLA_PUT_BE32(skb, CTA_SECMARK, htonl(ct->secmark));
255         return 0;
256
257 nla_put_failure:
258         return -1;
259 }
260 #else
261 #define ctnetlink_dump_secmark(a, b) (0)
262 #endif
263
264 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
265
266 static inline int
267 ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
268 {
269         struct nlattr *nest_parms;
270
271         if (!(ct->status & IPS_EXPECTED))
272                 return 0;
273
274         nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER | NLA_F_NESTED);
275         if (!nest_parms)
276                 goto nla_put_failure;
277         if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
278                 goto nla_put_failure;
279         nla_nest_end(skb, nest_parms);
280
281         return 0;
282
283 nla_put_failure:
284         return -1;
285 }
286
287 #ifdef CONFIG_NF_NAT_NEEDED
288 static int
289 dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
290 {
291         struct nlattr *nest_parms;
292
293         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
294         if (!nest_parms)
295                 goto nla_put_failure;
296
297         NLA_PUT_BE32(skb, CTA_NAT_SEQ_CORRECTION_POS,
298                      htonl(natseq->correction_pos));
299         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_BEFORE,
300                      htonl(natseq->offset_before));
301         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_AFTER,
302                      htonl(natseq->offset_after));
303
304         nla_nest_end(skb, nest_parms);
305
306         return 0;
307
308 nla_put_failure:
309         return -1;
310 }
311
312 static inline int
313 ctnetlink_dump_nat_seq_adj(struct sk_buff *skb, const struct nf_conn *ct)
314 {
315         struct nf_nat_seq *natseq;
316         struct nf_conn_nat *nat = nfct_nat(ct);
317
318         if (!(ct->status & IPS_SEQ_ADJUST) || !nat)
319                 return 0;
320
321         natseq = &nat->seq[IP_CT_DIR_ORIGINAL];
322         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_ORIG) == -1)
323                 return -1;
324
325         natseq = &nat->seq[IP_CT_DIR_REPLY];
326         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_REPLY) == -1)
327                 return -1;
328
329         return 0;
330 }
331 #else
332 #define ctnetlink_dump_nat_seq_adj(a, b) (0)
333 #endif
334
335 static inline int
336 ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
337 {
338         NLA_PUT_BE32(skb, CTA_ID, htonl((unsigned long)ct));
339         return 0;
340
341 nla_put_failure:
342         return -1;
343 }
344
345 static inline int
346 ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
347 {
348         NLA_PUT_BE32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use)));
349         return 0;
350
351 nla_put_failure:
352         return -1;
353 }
354
355 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
356
357 static int
358 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
359                     int event, int nowait,
360                     const struct nf_conn *ct)
361 {
362         struct nlmsghdr *nlh;
363         struct nfgenmsg *nfmsg;
364         struct nlattr *nest_parms;
365         unsigned char *b = skb_tail_pointer(skb);
366
367         event |= NFNL_SUBSYS_CTNETLINK << 8;
368         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
369         nfmsg  = NLMSG_DATA(nlh);
370
371         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
372         nfmsg->nfgen_family = nf_ct_l3num(ct);
373         nfmsg->version      = NFNETLINK_V0;
374         nfmsg->res_id       = 0;
375
376         nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
377         if (!nest_parms)
378                 goto nla_put_failure;
379         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
380                 goto nla_put_failure;
381         nla_nest_end(skb, nest_parms);
382
383         nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
384         if (!nest_parms)
385                 goto nla_put_failure;
386         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
387                 goto nla_put_failure;
388         nla_nest_end(skb, nest_parms);
389
390         if (ctnetlink_dump_status(skb, ct) < 0 ||
391             ctnetlink_dump_timeout(skb, ct) < 0 ||
392             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
393             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
394             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
395             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
396             ctnetlink_dump_mark(skb, ct) < 0 ||
397             ctnetlink_dump_secmark(skb, ct) < 0 ||
398             ctnetlink_dump_id(skb, ct) < 0 ||
399             ctnetlink_dump_use(skb, ct) < 0 ||
400             ctnetlink_dump_master(skb, ct) < 0 ||
401             ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
402                 goto nla_put_failure;
403
404         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
405         return skb->len;
406
407 nlmsg_failure:
408 nla_put_failure:
409         nlmsg_trim(skb, b);
410         return -1;
411 }
412
413 #ifdef CONFIG_NF_CONNTRACK_EVENTS
414 static int ctnetlink_conntrack_event(struct notifier_block *this,
415                                      unsigned long events, void *ptr)
416 {
417         struct nlmsghdr *nlh;
418         struct nfgenmsg *nfmsg;
419         struct nlattr *nest_parms;
420         struct nf_conn *ct = (struct nf_conn *)ptr;
421         struct sk_buff *skb;
422         unsigned int type;
423         sk_buff_data_t b;
424         unsigned int flags = 0, group;
425
426         /* ignore our fake conntrack entry */
427         if (ct == &nf_conntrack_untracked)
428                 return NOTIFY_DONE;
429
430         if (events & IPCT_DESTROY) {
431                 type = IPCTNL_MSG_CT_DELETE;
432                 group = NFNLGRP_CONNTRACK_DESTROY;
433         } else  if (events & (IPCT_NEW | IPCT_RELATED)) {
434                 type = IPCTNL_MSG_CT_NEW;
435                 flags = NLM_F_CREATE|NLM_F_EXCL;
436                 group = NFNLGRP_CONNTRACK_NEW;
437         } else  if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
438                 type = IPCTNL_MSG_CT_NEW;
439                 group = NFNLGRP_CONNTRACK_UPDATE;
440         } else
441                 return NOTIFY_DONE;
442
443         if (!nfnetlink_has_listeners(group))
444                 return NOTIFY_DONE;
445
446         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
447         if (!skb)
448                 return NOTIFY_DONE;
449
450         b = skb->tail;
451
452         type |= NFNL_SUBSYS_CTNETLINK << 8;
453         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
454         nfmsg = NLMSG_DATA(nlh);
455
456         nlh->nlmsg_flags    = flags;
457         nfmsg->nfgen_family = nf_ct_l3num(ct);
458         nfmsg->version  = NFNETLINK_V0;
459         nfmsg->res_id   = 0;
460
461         nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
462         if (!nest_parms)
463                 goto nla_put_failure;
464         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
465                 goto nla_put_failure;
466         nla_nest_end(skb, nest_parms);
467
468         nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
469         if (!nest_parms)
470                 goto nla_put_failure;
471         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
472                 goto nla_put_failure;
473         nla_nest_end(skb, nest_parms);
474
475         if (ctnetlink_dump_id(skb, ct) < 0)
476                 goto nla_put_failure;
477
478         if (ctnetlink_dump_status(skb, ct) < 0)
479                 goto nla_put_failure;
480
481         if (events & IPCT_DESTROY) {
482                 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
483                     ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
484                         goto nla_put_failure;
485         } else {
486                 if (ctnetlink_dump_timeout(skb, ct) < 0)
487                         goto nla_put_failure;
488
489                 if (events & IPCT_PROTOINFO
490                     && ctnetlink_dump_protoinfo(skb, ct) < 0)
491                         goto nla_put_failure;
492
493                 if ((events & IPCT_HELPER || nfct_help(ct))
494                     && ctnetlink_dump_helpinfo(skb, ct) < 0)
495                         goto nla_put_failure;
496
497 #ifdef CONFIG_NF_CONNTRACK_SECMARK
498                 if ((events & IPCT_SECMARK || ct->secmark)
499                     && ctnetlink_dump_secmark(skb, ct) < 0)
500                         goto nla_put_failure;
501 #endif
502
503                 if (events & IPCT_COUNTER_FILLING &&
504                     (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
505                      ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0))
506                         goto nla_put_failure;
507
508                 if (events & IPCT_RELATED &&
509                     ctnetlink_dump_master(skb, ct) < 0)
510                         goto nla_put_failure;
511
512                 if (events & IPCT_NATSEQADJ &&
513                     ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
514                         goto nla_put_failure;
515         }
516
517 #ifdef CONFIG_NF_CONNTRACK_MARK
518         if ((events & IPCT_MARK || ct->mark)
519             && ctnetlink_dump_mark(skb, ct) < 0)
520                 goto nla_put_failure;
521 #endif
522
523         nlh->nlmsg_len = skb->tail - b;
524         nfnetlink_send(skb, 0, group, 0);
525         return NOTIFY_DONE;
526
527 nlmsg_failure:
528 nla_put_failure:
529         kfree_skb(skb);
530         return NOTIFY_DONE;
531 }
532 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
533
534 static int ctnetlink_done(struct netlink_callback *cb)
535 {
536         if (cb->args[1])
537                 nf_ct_put((struct nf_conn *)cb->args[1]);
538         return 0;
539 }
540
541 static int
542 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
543 {
544         struct nf_conn *ct, *last;
545         struct nf_conntrack_tuple_hash *h;
546         struct hlist_node *n;
547         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
548         u_int8_t l3proto = nfmsg->nfgen_family;
549
550         rcu_read_lock();
551         last = (struct nf_conn *)cb->args[1];
552         for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
553 restart:
554                 hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[cb->args[0]],
555                                          hnode) {
556                         if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
557                                 continue;
558                         ct = nf_ct_tuplehash_to_ctrack(h);
559                         /* Dump entries of a given L3 protocol number.
560                          * If it is not specified, ie. l3proto == 0,
561                          * then dump everything. */
562                         if (l3proto && nf_ct_l3num(ct) != l3proto)
563                                 continue;
564                         if (cb->args[1]) {
565                                 if (ct != last)
566                                         continue;
567                                 cb->args[1] = 0;
568                         }
569                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
570                                                 cb->nlh->nlmsg_seq,
571                                                 IPCTNL_MSG_CT_NEW,
572                                                 1, ct) < 0) {
573                                 if (!atomic_inc_not_zero(&ct->ct_general.use))
574                                         continue;
575                                 cb->args[1] = (unsigned long)ct;
576                                 goto out;
577                         }
578 #ifdef CONFIG_NF_CT_ACCT
579                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
580                                                 IPCTNL_MSG_CT_GET_CTRZERO)
581                                 memset(&ct->counters, 0, sizeof(ct->counters));
582 #endif
583                 }
584                 if (cb->args[1]) {
585                         cb->args[1] = 0;
586                         goto restart;
587                 }
588         }
589 out:
590         rcu_read_unlock();
591         if (last)
592                 nf_ct_put(last);
593
594         return skb->len;
595 }
596
597 static inline int
598 ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
599 {
600         struct nlattr *tb[CTA_IP_MAX+1];
601         struct nf_conntrack_l3proto *l3proto;
602         int ret = 0;
603
604         nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
605
606         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
607
608         if (likely(l3proto->nlattr_to_tuple)) {
609                 ret = nla_validate_nested(attr, CTA_IP_MAX,
610                                           l3proto->nla_policy);
611                 if (ret == 0)
612                         ret = l3proto->nlattr_to_tuple(tb, tuple);
613         }
614
615         nf_ct_l3proto_put(l3proto);
616
617         return ret;
618 }
619
620 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
621         [CTA_PROTO_NUM] = { .type = NLA_U8 },
622 };
623
624 static inline int
625 ctnetlink_parse_tuple_proto(struct nlattr *attr,
626                             struct nf_conntrack_tuple *tuple)
627 {
628         struct nlattr *tb[CTA_PROTO_MAX+1];
629         struct nf_conntrack_l4proto *l4proto;
630         int ret = 0;
631
632         ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
633         if (ret < 0)
634                 return ret;
635
636         if (!tb[CTA_PROTO_NUM])
637                 return -EINVAL;
638         tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
639
640         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
641
642         if (likely(l4proto->nlattr_to_tuple)) {
643                 ret = nla_validate_nested(attr, CTA_PROTO_MAX,
644                                           l4proto->nla_policy);
645                 if (ret == 0)
646                         ret = l4proto->nlattr_to_tuple(tb, tuple);
647         }
648
649         nf_ct_l4proto_put(l4proto);
650
651         return ret;
652 }
653
654 static int
655 ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
656                       enum ctattr_tuple type, u_int8_t l3num)
657 {
658         struct nlattr *tb[CTA_TUPLE_MAX+1];
659         int err;
660
661         memset(tuple, 0, sizeof(*tuple));
662
663         nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
664
665         if (!tb[CTA_TUPLE_IP])
666                 return -EINVAL;
667
668         tuple->src.l3num = l3num;
669
670         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
671         if (err < 0)
672                 return err;
673
674         if (!tb[CTA_TUPLE_PROTO])
675                 return -EINVAL;
676
677         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
678         if (err < 0)
679                 return err;
680
681         /* orig and expect tuples get DIR_ORIGINAL */
682         if (type == CTA_TUPLE_REPLY)
683                 tuple->dst.dir = IP_CT_DIR_REPLY;
684         else
685                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
686
687         return 0;
688 }
689
690 #ifdef CONFIG_NF_NAT_NEEDED
691 static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = {
692         [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 },
693         [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 },
694 };
695
696 static int nfnetlink_parse_nat_proto(struct nlattr *attr,
697                                      const struct nf_conn *ct,
698                                      struct nf_nat_range *range)
699 {
700         struct nlattr *tb[CTA_PROTONAT_MAX+1];
701         const struct nf_nat_protocol *npt;
702         int err;
703
704         err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy);
705         if (err < 0)
706                 return err;
707
708         npt = nf_nat_proto_find_get(nf_ct_protonum(ct));
709         if (npt->nlattr_to_range)
710                 err = npt->nlattr_to_range(tb, range);
711         nf_nat_proto_put(npt);
712         return err;
713 }
714
715 static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = {
716         [CTA_NAT_MINIP]         = { .type = NLA_U32 },
717         [CTA_NAT_MAXIP]         = { .type = NLA_U32 },
718 };
719
720 static inline int
721 nfnetlink_parse_nat(struct nlattr *nat,
722                     const struct nf_conn *ct, struct nf_nat_range *range)
723 {
724         struct nlattr *tb[CTA_NAT_MAX+1];
725         int err;
726
727         memset(range, 0, sizeof(*range));
728
729         err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy);
730         if (err < 0)
731                 return err;
732
733         if (tb[CTA_NAT_MINIP])
734                 range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]);
735
736         if (!tb[CTA_NAT_MAXIP])
737                 range->max_ip = range->min_ip;
738         else
739                 range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]);
740
741         if (range->min_ip)
742                 range->flags |= IP_NAT_RANGE_MAP_IPS;
743
744         if (!tb[CTA_NAT_PROTO])
745                 return 0;
746
747         err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range);
748         if (err < 0)
749                 return err;
750
751         return 0;
752 }
753 #endif
754
755 static inline int
756 ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
757 {
758         struct nlattr *tb[CTA_HELP_MAX+1];
759
760         nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
761
762         if (!tb[CTA_HELP_NAME])
763                 return -EINVAL;
764
765         *helper_name = nla_data(tb[CTA_HELP_NAME]);
766
767         return 0;
768 }
769
770 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
771         [CTA_STATUS]            = { .type = NLA_U32 },
772         [CTA_TIMEOUT]           = { .type = NLA_U32 },
773         [CTA_MARK]              = { .type = NLA_U32 },
774         [CTA_USE]               = { .type = NLA_U32 },
775         [CTA_ID]                = { .type = NLA_U32 },
776 };
777
778 static int
779 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
780                         struct nlmsghdr *nlh, struct nlattr *cda[])
781 {
782         struct nf_conntrack_tuple_hash *h;
783         struct nf_conntrack_tuple tuple;
784         struct nf_conn *ct;
785         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
786         u_int8_t u3 = nfmsg->nfgen_family;
787         int err = 0;
788
789         if (cda[CTA_TUPLE_ORIG])
790                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
791         else if (cda[CTA_TUPLE_REPLY])
792                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
793         else {
794                 /* Flush the whole table */
795                 nf_conntrack_flush();
796                 return 0;
797         }
798
799         if (err < 0)
800                 return err;
801
802         h = nf_conntrack_find_get(&tuple);
803         if (!h)
804                 return -ENOENT;
805
806         ct = nf_ct_tuplehash_to_ctrack(h);
807
808         if (cda[CTA_ID]) {
809                 u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
810                 if (id != (u32)(unsigned long)ct) {
811                         nf_ct_put(ct);
812                         return -ENOENT;
813                 }
814         }
815
816         nf_ct_kill(ct);
817         nf_ct_put(ct);
818
819         return 0;
820 }
821
822 static int
823 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
824                         struct nlmsghdr *nlh, struct nlattr *cda[])
825 {
826         struct nf_conntrack_tuple_hash *h;
827         struct nf_conntrack_tuple tuple;
828         struct nf_conn *ct;
829         struct sk_buff *skb2 = NULL;
830         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
831         u_int8_t u3 = nfmsg->nfgen_family;
832         int err = 0;
833
834         if (nlh->nlmsg_flags & NLM_F_DUMP) {
835 #ifndef CONFIG_NF_CT_ACCT
836                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
837                         return -ENOTSUPP;
838 #endif
839                 return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
840                                           ctnetlink_done);
841         }
842
843         if (cda[CTA_TUPLE_ORIG])
844                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
845         else if (cda[CTA_TUPLE_REPLY])
846                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
847         else
848                 return -EINVAL;
849
850         if (err < 0)
851                 return err;
852
853         h = nf_conntrack_find_get(&tuple);
854         if (!h)
855                 return -ENOENT;
856
857         ct = nf_ct_tuplehash_to_ctrack(h);
858
859         err = -ENOMEM;
860         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
861         if (!skb2) {
862                 nf_ct_put(ct);
863                 return -ENOMEM;
864         }
865
866         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
867                                   IPCTNL_MSG_CT_NEW, 1, ct);
868         nf_ct_put(ct);
869         if (err <= 0)
870                 goto free;
871
872         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
873         if (err < 0)
874                 goto out;
875
876         return 0;
877
878 free:
879         kfree_skb(skb2);
880 out:
881         return err;
882 }
883
884 static int
885 ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
886 {
887         unsigned long d;
888         unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
889         d = ct->status ^ status;
890
891         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
892                 /* unchangeable */
893                 return -EBUSY;
894
895         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
896                 /* SEEN_REPLY bit can only be set */
897                 return -EBUSY;
898
899         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
900                 /* ASSURED bit can only be set */
901                 return -EBUSY;
902
903         if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
904 #ifndef CONFIG_NF_NAT_NEEDED
905                 return -EOPNOTSUPP;
906 #else
907                 struct nf_nat_range range;
908
909                 if (cda[CTA_NAT_DST]) {
910                         if (nfnetlink_parse_nat(cda[CTA_NAT_DST], ct,
911                                                 &range) < 0)
912                                 return -EINVAL;
913                         if (nf_nat_initialized(ct, IP_NAT_MANIP_DST))
914                                 return -EEXIST;
915                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
916                 }
917                 if (cda[CTA_NAT_SRC]) {
918                         if (nfnetlink_parse_nat(cda[CTA_NAT_SRC], ct,
919                                                 &range) < 0)
920                                 return -EINVAL;
921                         if (nf_nat_initialized(ct, IP_NAT_MANIP_SRC))
922                                 return -EEXIST;
923                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
924                 }
925 #endif
926         }
927
928         /* Be careful here, modifying NAT bits can screw up things,
929          * so don't let users modify them directly if they don't pass
930          * nf_nat_range. */
931         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
932         return 0;
933 }
934
935
936 static inline int
937 ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
938 {
939         struct nf_conntrack_helper *helper;
940         struct nf_conn_help *help = nfct_help(ct);
941         char *helpname;
942         int err;
943
944         /* don't change helper of sibling connections */
945         if (ct->master)
946                 return -EBUSY;
947
948         err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
949         if (err < 0)
950                 return err;
951
952         if (!strcmp(helpname, "")) {
953                 if (help && help->helper) {
954                         /* we had a helper before ... */
955                         nf_ct_remove_expectations(ct);
956                         rcu_assign_pointer(help->helper, NULL);
957                 }
958
959                 return 0;
960         }
961
962         helper = __nf_conntrack_helper_find_byname(helpname);
963         if (helper == NULL)
964                 return -EOPNOTSUPP;
965
966         if (help) {
967                 if (help->helper == helper)
968                         return 0;
969                 if (help->helper)
970                         return -EBUSY;
971                 /* need to zero data of old helper */
972                 memset(&help->help, 0, sizeof(help->help));
973         } else {
974                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
975                 if (help == NULL)
976                         return -ENOMEM;
977         }
978
979         rcu_assign_pointer(help->helper, helper);
980
981         return 0;
982 }
983
984 static inline int
985 ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
986 {
987         u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
988
989         if (!del_timer(&ct->timeout))
990                 return -ETIME;
991
992         ct->timeout.expires = jiffies + timeout * HZ;
993         add_timer(&ct->timeout);
994
995         return 0;
996 }
997
998 static inline int
999 ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
1000 {
1001         struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
1002         struct nf_conntrack_l4proto *l4proto;
1003         int err = 0;
1004
1005         nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1006
1007         l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
1008         if (l4proto->from_nlattr)
1009                 err = l4proto->from_nlattr(tb, ct);
1010         nf_ct_l4proto_put(l4proto);
1011
1012         return err;
1013 }
1014
1015 #ifdef CONFIG_NF_NAT_NEEDED
1016 static inline int
1017 change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1018 {
1019         struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1020
1021         nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1022
1023         if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1024                 return -EINVAL;
1025
1026         natseq->correction_pos =
1027                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1028
1029         if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1030                 return -EINVAL;
1031
1032         natseq->offset_before =
1033                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1034
1035         if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1036                 return -EINVAL;
1037
1038         natseq->offset_after =
1039                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1040
1041         return 0;
1042 }
1043
1044 static int
1045 ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1046 {
1047         int ret = 0;
1048         struct nf_conn_nat *nat = nfct_nat(ct);
1049
1050         if (!nat)
1051                 return 0;
1052
1053         if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1054                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1055                                          cda[CTA_NAT_SEQ_ADJ_ORIG]);
1056                 if (ret < 0)
1057                         return ret;
1058
1059                 ct->status |= IPS_SEQ_ADJUST;
1060         }
1061
1062         if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1063                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1064                                          cda[CTA_NAT_SEQ_ADJ_REPLY]);
1065                 if (ret < 0)
1066                         return ret;
1067
1068                 ct->status |= IPS_SEQ_ADJUST;
1069         }
1070
1071         return 0;
1072 }
1073 #endif
1074
1075 static int
1076 ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1077 {
1078         int err;
1079
1080         if (cda[CTA_HELP]) {
1081                 err = ctnetlink_change_helper(ct, cda);
1082                 if (err < 0)
1083                         return err;
1084         }
1085
1086         if (cda[CTA_TIMEOUT]) {
1087                 err = ctnetlink_change_timeout(ct, cda);
1088                 if (err < 0)
1089                         return err;
1090         }
1091
1092         if (cda[CTA_STATUS]) {
1093                 err = ctnetlink_change_status(ct, cda);
1094                 if (err < 0)
1095                         return err;
1096         }
1097
1098         if (cda[CTA_PROTOINFO]) {
1099                 err = ctnetlink_change_protoinfo(ct, cda);
1100                 if (err < 0)
1101                         return err;
1102         }
1103
1104 #if defined(CONFIG_NF_CONNTRACK_MARK)
1105         if (cda[CTA_MARK])
1106                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1107 #endif
1108
1109 #ifdef CONFIG_NF_NAT_NEEDED
1110         if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1111                 err = ctnetlink_change_nat_seq_adj(ct, cda);
1112                 if (err < 0)
1113                         return err;
1114         }
1115 #endif
1116
1117         return 0;
1118 }
1119
1120 static int
1121 ctnetlink_create_conntrack(struct nlattr *cda[],
1122                            struct nf_conntrack_tuple *otuple,
1123                            struct nf_conntrack_tuple *rtuple,
1124                            struct nf_conn *master_ct)
1125 {
1126         struct nf_conn *ct;
1127         int err = -EINVAL;
1128         struct nf_conn_help *help;
1129         struct nf_conntrack_helper *helper;
1130
1131         ct = nf_conntrack_alloc(otuple, rtuple, GFP_KERNEL);
1132         if (ct == NULL || IS_ERR(ct))
1133                 return -ENOMEM;
1134
1135         if (!cda[CTA_TIMEOUT])
1136                 goto err;
1137         ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1138
1139         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1140         ct->status |= IPS_CONFIRMED;
1141
1142         if (cda[CTA_STATUS]) {
1143                 err = ctnetlink_change_status(ct, cda);
1144                 if (err < 0)
1145                         goto err;
1146         }
1147
1148         if (cda[CTA_PROTOINFO]) {
1149                 err = ctnetlink_change_protoinfo(ct, cda);
1150                 if (err < 0)
1151                         goto err;
1152         }
1153
1154 #if defined(CONFIG_NF_CONNTRACK_MARK)
1155         if (cda[CTA_MARK])
1156                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1157 #endif
1158
1159         rcu_read_lock();
1160         helper = __nf_ct_helper_find(rtuple);
1161         if (helper) {
1162                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
1163                 if (help == NULL) {
1164                         rcu_read_unlock();
1165                         err = -ENOMEM;
1166                         goto err;
1167                 }
1168                 /* not in hash table yet so not strictly necessary */
1169                 rcu_assign_pointer(help->helper, helper);
1170         }
1171
1172         /* setup master conntrack: this is a confirmed expectation */
1173         if (master_ct) {
1174                 __set_bit(IPS_EXPECTED_BIT, &ct->status);
1175                 ct->master = master_ct;
1176         }
1177
1178         add_timer(&ct->timeout);
1179         nf_conntrack_hash_insert(ct);
1180         rcu_read_unlock();
1181
1182         return 0;
1183
1184 err:
1185         nf_conntrack_free(ct);
1186         return err;
1187 }
1188
1189 static int
1190 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1191                         struct nlmsghdr *nlh, struct nlattr *cda[])
1192 {
1193         struct nf_conntrack_tuple otuple, rtuple;
1194         struct nf_conntrack_tuple_hash *h = NULL;
1195         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1196         u_int8_t u3 = nfmsg->nfgen_family;
1197         int err = 0;
1198
1199         if (cda[CTA_TUPLE_ORIG]) {
1200                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1201                 if (err < 0)
1202                         return err;
1203         }
1204
1205         if (cda[CTA_TUPLE_REPLY]) {
1206                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1207                 if (err < 0)
1208                         return err;
1209         }
1210
1211         spin_lock_bh(&nf_conntrack_lock);
1212         if (cda[CTA_TUPLE_ORIG])
1213                 h = __nf_conntrack_find(&otuple);
1214         else if (cda[CTA_TUPLE_REPLY])
1215                 h = __nf_conntrack_find(&rtuple);
1216
1217         if (h == NULL) {
1218                 struct nf_conntrack_tuple master;
1219                 struct nf_conntrack_tuple_hash *master_h = NULL;
1220                 struct nf_conn *master_ct = NULL;
1221
1222                 if (cda[CTA_TUPLE_MASTER]) {
1223                         err = ctnetlink_parse_tuple(cda,
1224                                                     &master,
1225                                                     CTA_TUPLE_MASTER,
1226                                                     u3);
1227                         if (err < 0)
1228                                 goto out_unlock;
1229
1230                         master_h = __nf_conntrack_find(&master);
1231                         if (master_h == NULL) {
1232                                 err = -ENOENT;
1233                                 goto out_unlock;
1234                         }
1235                         master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1236                         atomic_inc(&master_ct->ct_general.use);
1237                 }
1238
1239                 spin_unlock_bh(&nf_conntrack_lock);
1240                 err = -ENOENT;
1241                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1242                         err = ctnetlink_create_conntrack(cda,
1243                                                          &otuple,
1244                                                          &rtuple,
1245                                                          master_ct);
1246                 if (err < 0 && master_ct)
1247                         nf_ct_put(master_ct);
1248
1249                 return err;
1250         }
1251         /* implicit 'else' */
1252
1253         /* We manipulate the conntrack inside the global conntrack table lock,
1254          * so there's no need to increase the refcount */
1255         err = -EEXIST;
1256         if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1257                 /* we only allow nat config for new conntracks */
1258                 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1259                         err = -EOPNOTSUPP;
1260                         goto out_unlock;
1261                 }
1262                 /* can't link an existing conntrack to a master */
1263                 if (cda[CTA_TUPLE_MASTER]) {
1264                         err = -EOPNOTSUPP;
1265                         goto out_unlock;
1266                 }
1267                 err = ctnetlink_change_conntrack(nf_ct_tuplehash_to_ctrack(h),
1268                                                  cda);
1269         }
1270
1271 out_unlock:
1272         spin_unlock_bh(&nf_conntrack_lock);
1273         return err;
1274 }
1275
1276 /***********************************************************************
1277  * EXPECT
1278  ***********************************************************************/
1279
1280 static inline int
1281 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1282                          const struct nf_conntrack_tuple *tuple,
1283                          enum ctattr_expect type)
1284 {
1285         struct nlattr *nest_parms;
1286
1287         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1288         if (!nest_parms)
1289                 goto nla_put_failure;
1290         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1291                 goto nla_put_failure;
1292         nla_nest_end(skb, nest_parms);
1293
1294         return 0;
1295
1296 nla_put_failure:
1297         return -1;
1298 }
1299
1300 static inline int
1301 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1302                         const struct nf_conntrack_tuple *tuple,
1303                         const struct nf_conntrack_tuple_mask *mask)
1304 {
1305         int ret;
1306         struct nf_conntrack_l3proto *l3proto;
1307         struct nf_conntrack_l4proto *l4proto;
1308         struct nf_conntrack_tuple m;
1309         struct nlattr *nest_parms;
1310
1311         memset(&m, 0xFF, sizeof(m));
1312         m.src.u.all = mask->src.u.all;
1313         memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1314
1315         nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1316         if (!nest_parms)
1317                 goto nla_put_failure;
1318
1319         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
1320         ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1321         nf_ct_l3proto_put(l3proto);
1322
1323         if (unlikely(ret < 0))
1324                 goto nla_put_failure;
1325
1326         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
1327         ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1328         nf_ct_l4proto_put(l4proto);
1329         if (unlikely(ret < 0))
1330                 goto nla_put_failure;
1331
1332         nla_nest_end(skb, nest_parms);
1333
1334         return 0;
1335
1336 nla_put_failure:
1337         return -1;
1338 }
1339
1340 static int
1341 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1342                           const struct nf_conntrack_expect *exp)
1343 {
1344         struct nf_conn *master = exp->master;
1345         long timeout = (exp->timeout.expires - jiffies) / HZ;
1346
1347         if (timeout < 0)
1348                 timeout = 0;
1349
1350         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1351                 goto nla_put_failure;
1352         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1353                 goto nla_put_failure;
1354         if (ctnetlink_exp_dump_tuple(skb,
1355                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1356                                  CTA_EXPECT_MASTER) < 0)
1357                 goto nla_put_failure;
1358
1359         NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1360         NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1361
1362         return 0;
1363
1364 nla_put_failure:
1365         return -1;
1366 }
1367
1368 static int
1369 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1370                     int event,
1371                     int nowait,
1372                     const struct nf_conntrack_expect *exp)
1373 {
1374         struct nlmsghdr *nlh;
1375         struct nfgenmsg *nfmsg;
1376         unsigned char *b = skb_tail_pointer(skb);
1377
1378         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1379         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1380         nfmsg  = NLMSG_DATA(nlh);
1381
1382         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1383         nfmsg->nfgen_family = exp->tuple.src.l3num;
1384         nfmsg->version      = NFNETLINK_V0;
1385         nfmsg->res_id       = 0;
1386
1387         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1388                 goto nla_put_failure;
1389
1390         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1391         return skb->len;
1392
1393 nlmsg_failure:
1394 nla_put_failure:
1395         nlmsg_trim(skb, b);
1396         return -1;
1397 }
1398
1399 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1400 static int ctnetlink_expect_event(struct notifier_block *this,
1401                                   unsigned long events, void *ptr)
1402 {
1403         struct nlmsghdr *nlh;
1404         struct nfgenmsg *nfmsg;
1405         struct nf_conntrack_expect *exp = (struct nf_conntrack_expect *)ptr;
1406         struct sk_buff *skb;
1407         unsigned int type;
1408         sk_buff_data_t b;
1409         int flags = 0;
1410
1411         if (events & IPEXP_NEW) {
1412                 type = IPCTNL_MSG_EXP_NEW;
1413                 flags = NLM_F_CREATE|NLM_F_EXCL;
1414         } else
1415                 return NOTIFY_DONE;
1416
1417         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1418                 return NOTIFY_DONE;
1419
1420         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1421         if (!skb)
1422                 return NOTIFY_DONE;
1423
1424         b = skb->tail;
1425
1426         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1427         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1428         nfmsg = NLMSG_DATA(nlh);
1429
1430         nlh->nlmsg_flags    = flags;
1431         nfmsg->nfgen_family = exp->tuple.src.l3num;
1432         nfmsg->version      = NFNETLINK_V0;
1433         nfmsg->res_id       = 0;
1434
1435         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1436                 goto nla_put_failure;
1437
1438         nlh->nlmsg_len = skb->tail - b;
1439         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1440         return NOTIFY_DONE;
1441
1442 nlmsg_failure:
1443 nla_put_failure:
1444         kfree_skb(skb);
1445         return NOTIFY_DONE;
1446 }
1447 #endif
1448 static int ctnetlink_exp_done(struct netlink_callback *cb)
1449 {
1450         if (cb->args[1])
1451                 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1452         return 0;
1453 }
1454
1455 static int
1456 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1457 {
1458         struct nf_conntrack_expect *exp, *last;
1459         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
1460         struct hlist_node *n;
1461         u_int8_t l3proto = nfmsg->nfgen_family;
1462
1463         rcu_read_lock();
1464         last = (struct nf_conntrack_expect *)cb->args[1];
1465         for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1466 restart:
1467                 hlist_for_each_entry(exp, n, &nf_ct_expect_hash[cb->args[0]],
1468                                      hnode) {
1469                         if (l3proto && exp->tuple.src.l3num != l3proto)
1470                                 continue;
1471                         if (cb->args[1]) {
1472                                 if (exp != last)
1473                                         continue;
1474                                 cb->args[1] = 0;
1475                         }
1476                         if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1477                                                     cb->nlh->nlmsg_seq,
1478                                                     IPCTNL_MSG_EXP_NEW,
1479                                                     1, exp) < 0) {
1480                                 if (!atomic_inc_not_zero(&exp->use))
1481                                         continue;
1482                                 cb->args[1] = (unsigned long)exp;
1483                                 goto out;
1484                         }
1485                 }
1486                 if (cb->args[1]) {
1487                         cb->args[1] = 0;
1488                         goto restart;
1489                 }
1490         }
1491 out:
1492         rcu_read_unlock();
1493         if (last)
1494                 nf_ct_expect_put(last);
1495
1496         return skb->len;
1497 }
1498
1499 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1500         [CTA_EXPECT_TIMEOUT]    = { .type = NLA_U32 },
1501         [CTA_EXPECT_ID]         = { .type = NLA_U32 },
1502 };
1503
1504 static int
1505 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1506                      struct nlmsghdr *nlh, struct nlattr *cda[])
1507 {
1508         struct nf_conntrack_tuple tuple;
1509         struct nf_conntrack_expect *exp;
1510         struct sk_buff *skb2;
1511         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1512         u_int8_t u3 = nfmsg->nfgen_family;
1513         int err = 0;
1514
1515         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1516                 return netlink_dump_start(ctnl, skb, nlh,
1517                                           ctnetlink_exp_dump_table,
1518                                           ctnetlink_exp_done);
1519         }
1520
1521         if (cda[CTA_EXPECT_MASTER])
1522                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1523         else
1524                 return -EINVAL;
1525
1526         if (err < 0)
1527                 return err;
1528
1529         exp = nf_ct_expect_find_get(&tuple);
1530         if (!exp)
1531                 return -ENOENT;
1532
1533         if (cda[CTA_EXPECT_ID]) {
1534                 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1535                 if (ntohl(id) != (u32)(unsigned long)exp) {
1536                         nf_ct_expect_put(exp);
1537                         return -ENOENT;
1538                 }
1539         }
1540
1541         err = -ENOMEM;
1542         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1543         if (!skb2)
1544                 goto out;
1545
1546         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1547                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1548                                       1, exp);
1549         if (err <= 0)
1550                 goto free;
1551
1552         nf_ct_expect_put(exp);
1553
1554         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1555
1556 free:
1557         kfree_skb(skb2);
1558 out:
1559         nf_ct_expect_put(exp);
1560         return err;
1561 }
1562
1563 static int
1564 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1565                      struct nlmsghdr *nlh, struct nlattr *cda[])
1566 {
1567         struct nf_conntrack_expect *exp;
1568         struct nf_conntrack_tuple tuple;
1569         struct nf_conntrack_helper *h;
1570         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1571         struct hlist_node *n, *next;
1572         u_int8_t u3 = nfmsg->nfgen_family;
1573         unsigned int i;
1574         int err;
1575
1576         if (cda[CTA_EXPECT_TUPLE]) {
1577                 /* delete a single expect by tuple */
1578                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1579                 if (err < 0)
1580                         return err;
1581
1582                 /* bump usage count to 2 */
1583                 exp = nf_ct_expect_find_get(&tuple);
1584                 if (!exp)
1585                         return -ENOENT;
1586
1587                 if (cda[CTA_EXPECT_ID]) {
1588                         __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1589                         if (ntohl(id) != (u32)(unsigned long)exp) {
1590                                 nf_ct_expect_put(exp);
1591                                 return -ENOENT;
1592                         }
1593                 }
1594
1595                 /* after list removal, usage count == 1 */
1596                 nf_ct_unexpect_related(exp);
1597                 /* have to put what we 'get' above.
1598                  * after this line usage count == 0 */
1599                 nf_ct_expect_put(exp);
1600         } else if (cda[CTA_EXPECT_HELP_NAME]) {
1601                 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1602                 struct nf_conn_help *m_help;
1603
1604                 /* delete all expectations for this helper */
1605                 spin_lock_bh(&nf_conntrack_lock);
1606                 h = __nf_conntrack_helper_find_byname(name);
1607                 if (!h) {
1608                         spin_unlock_bh(&nf_conntrack_lock);
1609                         return -EOPNOTSUPP;
1610                 }
1611                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1612                         hlist_for_each_entry_safe(exp, n, next,
1613                                                   &nf_ct_expect_hash[i],
1614                                                   hnode) {
1615                                 m_help = nfct_help(exp->master);
1616                                 if (m_help->helper == h
1617                                     && del_timer(&exp->timeout)) {
1618                                         nf_ct_unlink_expect(exp);
1619                                         nf_ct_expect_put(exp);
1620                                 }
1621                         }
1622                 }
1623                 spin_unlock_bh(&nf_conntrack_lock);
1624         } else {
1625                 /* This basically means we have to flush everything*/
1626                 spin_lock_bh(&nf_conntrack_lock);
1627                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1628                         hlist_for_each_entry_safe(exp, n, next,
1629                                                   &nf_ct_expect_hash[i],
1630                                                   hnode) {
1631                                 if (del_timer(&exp->timeout)) {
1632                                         nf_ct_unlink_expect(exp);
1633                                         nf_ct_expect_put(exp);
1634                                 }
1635                         }
1636                 }
1637                 spin_unlock_bh(&nf_conntrack_lock);
1638         }
1639
1640         return 0;
1641 }
1642 static int
1643 ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1644 {
1645         return -EOPNOTSUPP;
1646 }
1647
1648 static int
1649 ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
1650 {
1651         struct nf_conntrack_tuple tuple, mask, master_tuple;
1652         struct nf_conntrack_tuple_hash *h = NULL;
1653         struct nf_conntrack_expect *exp;
1654         struct nf_conn *ct;
1655         struct nf_conn_help *help;
1656         int err = 0;
1657
1658         /* caller guarantees that those three CTA_EXPECT_* exist */
1659         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1660         if (err < 0)
1661                 return err;
1662         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1663         if (err < 0)
1664                 return err;
1665         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1666         if (err < 0)
1667                 return err;
1668
1669         /* Look for master conntrack of this expectation */
1670         h = nf_conntrack_find_get(&master_tuple);
1671         if (!h)
1672                 return -ENOENT;
1673         ct = nf_ct_tuplehash_to_ctrack(h);
1674         help = nfct_help(ct);
1675
1676         if (!help || !help->helper) {
1677                 /* such conntrack hasn't got any helper, abort */
1678                 err = -EINVAL;
1679                 goto out;
1680         }
1681
1682         exp = nf_ct_expect_alloc(ct);
1683         if (!exp) {
1684                 err = -ENOMEM;
1685                 goto out;
1686         }
1687
1688         exp->expectfn = NULL;
1689         exp->flags = 0;
1690         exp->master = ct;
1691         exp->helper = NULL;
1692         memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1693         memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1694         exp->mask.src.u.all = mask.src.u.all;
1695
1696         err = nf_ct_expect_related(exp);
1697         nf_ct_expect_put(exp);
1698
1699 out:
1700         nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1701         return err;
1702 }
1703
1704 static int
1705 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1706                      struct nlmsghdr *nlh, struct nlattr *cda[])
1707 {
1708         struct nf_conntrack_tuple tuple;
1709         struct nf_conntrack_expect *exp;
1710         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1711         u_int8_t u3 = nfmsg->nfgen_family;
1712         int err = 0;
1713
1714         if (!cda[CTA_EXPECT_TUPLE]
1715             || !cda[CTA_EXPECT_MASK]
1716             || !cda[CTA_EXPECT_MASTER])
1717                 return -EINVAL;
1718
1719         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1720         if (err < 0)
1721                 return err;
1722
1723         spin_lock_bh(&nf_conntrack_lock);
1724         exp = __nf_ct_expect_find(&tuple);
1725
1726         if (!exp) {
1727                 spin_unlock_bh(&nf_conntrack_lock);
1728                 err = -ENOENT;
1729                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1730                         err = ctnetlink_create_expect(cda, u3);
1731                 return err;
1732         }
1733
1734         err = -EEXIST;
1735         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1736                 err = ctnetlink_change_expect(exp, cda);
1737         spin_unlock_bh(&nf_conntrack_lock);
1738
1739         return err;
1740 }
1741
1742 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1743 static struct notifier_block ctnl_notifier = {
1744         .notifier_call  = ctnetlink_conntrack_event,
1745 };
1746
1747 static struct notifier_block ctnl_notifier_exp = {
1748         .notifier_call  = ctnetlink_expect_event,
1749 };
1750 #endif
1751
1752 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1753         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1754                                             .attr_count = CTA_MAX,
1755                                             .policy = ct_nla_policy },
1756         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1757                                             .attr_count = CTA_MAX,
1758                                             .policy = ct_nla_policy },
1759         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1760                                             .attr_count = CTA_MAX,
1761                                             .policy = ct_nla_policy },
1762         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1763                                             .attr_count = CTA_MAX,
1764                                             .policy = ct_nla_policy },
1765 };
1766
1767 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1768         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1769                                             .attr_count = CTA_EXPECT_MAX,
1770                                             .policy = exp_nla_policy },
1771         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1772                                             .attr_count = CTA_EXPECT_MAX,
1773                                             .policy = exp_nla_policy },
1774         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1775                                             .attr_count = CTA_EXPECT_MAX,
1776                                             .policy = exp_nla_policy },
1777 };
1778
1779 static const struct nfnetlink_subsystem ctnl_subsys = {
1780         .name                           = "conntrack",
1781         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1782         .cb_count                       = IPCTNL_MSG_MAX,
1783         .cb                             = ctnl_cb,
1784 };
1785
1786 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1787         .name                           = "conntrack_expect",
1788         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1789         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1790         .cb                             = ctnl_exp_cb,
1791 };
1792
1793 MODULE_ALIAS("ip_conntrack_netlink");
1794 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1795 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1796
1797 static int __init ctnetlink_init(void)
1798 {
1799         int ret;
1800
1801         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1802         ret = nfnetlink_subsys_register(&ctnl_subsys);
1803         if (ret < 0) {
1804                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1805                 goto err_out;
1806         }
1807
1808         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1809         if (ret < 0) {
1810                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1811                 goto err_unreg_subsys;
1812         }
1813
1814 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1815         ret = nf_conntrack_register_notifier(&ctnl_notifier);
1816         if (ret < 0) {
1817                 printk("ctnetlink_init: cannot register notifier.\n");
1818                 goto err_unreg_exp_subsys;
1819         }
1820
1821         ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1822         if (ret < 0) {
1823                 printk("ctnetlink_init: cannot expect register notifier.\n");
1824                 goto err_unreg_notifier;
1825         }
1826 #endif
1827
1828         return 0;
1829
1830 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1831 err_unreg_notifier:
1832         nf_conntrack_unregister_notifier(&ctnl_notifier);
1833 err_unreg_exp_subsys:
1834         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1835 #endif
1836 err_unreg_subsys:
1837         nfnetlink_subsys_unregister(&ctnl_subsys);
1838 err_out:
1839         return ret;
1840 }
1841
1842 static void __exit ctnetlink_exit(void)
1843 {
1844         printk("ctnetlink: unregistering from nfnetlink.\n");
1845
1846 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1847         nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
1848         nf_conntrack_unregister_notifier(&ctnl_notifier);
1849 #endif
1850
1851         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1852         nfnetlink_subsys_unregister(&ctnl_subsys);
1853         return;
1854 }
1855
1856 module_init(ctnetlink_init);
1857 module_exit(ctnetlink_exit);