]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c
net: Convert uses of typedef ctl_table to struct ctl_table
[karo-tx-linux.git] / net / ipv4 / netfilter / nf_conntrack_l3proto_ipv4.c
1
2 /* (C) 1999-2001 Paul `Rusty' Russell
3  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
4  * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License version 2 as
8  * published by the Free Software Foundation.
9  */
10
11 #include <linux/types.h>
12 #include <linux/ip.h>
13 #include <linux/netfilter.h>
14 #include <linux/module.h>
15 #include <linux/skbuff.h>
16 #include <linux/icmp.h>
17 #include <linux/sysctl.h>
18 #include <net/route.h>
19 #include <net/ip.h>
20
21 #include <linux/netfilter_ipv4.h>
22 #include <net/netfilter/nf_conntrack.h>
23 #include <net/netfilter/nf_conntrack_helper.h>
24 #include <net/netfilter/nf_conntrack_l4proto.h>
25 #include <net/netfilter/nf_conntrack_l3proto.h>
26 #include <net/netfilter/nf_conntrack_zones.h>
27 #include <net/netfilter/nf_conntrack_core.h>
28 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
29 #include <net/netfilter/nf_nat_helper.h>
30 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
31 #include <net/netfilter/nf_log.h>
32
33 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
34                               struct nf_conntrack_tuple *tuple)
35 {
36         const __be32 *ap;
37         __be32 _addrs[2];
38         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
39                                 sizeof(u_int32_t) * 2, _addrs);
40         if (ap == NULL)
41                 return false;
42
43         tuple->src.u3.ip = ap[0];
44         tuple->dst.u3.ip = ap[1];
45
46         return true;
47 }
48
49 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
50                               const struct nf_conntrack_tuple *orig)
51 {
52         tuple->src.u3.ip = orig->dst.u3.ip;
53         tuple->dst.u3.ip = orig->src.u3.ip;
54
55         return true;
56 }
57
58 static int ipv4_print_tuple(struct seq_file *s,
59                             const struct nf_conntrack_tuple *tuple)
60 {
61         return seq_printf(s, "src=%pI4 dst=%pI4 ",
62                           &tuple->src.u3.ip, &tuple->dst.u3.ip);
63 }
64
65 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
66                             unsigned int *dataoff, u_int8_t *protonum)
67 {
68         const struct iphdr *iph;
69         struct iphdr _iph;
70
71         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
72         if (iph == NULL)
73                 return -NF_ACCEPT;
74
75         /* Conntrack defragments packets, we might still see fragments
76          * inside ICMP packets though. */
77         if (iph->frag_off & htons(IP_OFFSET))
78                 return -NF_ACCEPT;
79
80         *dataoff = nhoff + (iph->ihl << 2);
81         *protonum = iph->protocol;
82
83         /* Check bogus IP headers */
84         if (*dataoff > skb->len) {
85                 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
86                          "nhoff %u, ihl %u, skblen %u\n",
87                          nhoff, iph->ihl << 2, skb->len);
88                 return -NF_ACCEPT;
89         }
90
91         return NF_ACCEPT;
92 }
93
94 static unsigned int ipv4_helper(unsigned int hooknum,
95                                 struct sk_buff *skb,
96                                 const struct net_device *in,
97                                 const struct net_device *out,
98                                 int (*okfn)(struct sk_buff *))
99 {
100         struct nf_conn *ct;
101         enum ip_conntrack_info ctinfo;
102         const struct nf_conn_help *help;
103         const struct nf_conntrack_helper *helper;
104
105         /* This is where we call the helper: as the packet goes out. */
106         ct = nf_ct_get(skb, &ctinfo);
107         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
108                 return NF_ACCEPT;
109
110         help = nfct_help(ct);
111         if (!help)
112                 return NF_ACCEPT;
113
114         /* rcu_read_lock()ed by nf_hook_slow */
115         helper = rcu_dereference(help->helper);
116         if (!helper)
117                 return NF_ACCEPT;
118
119         return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
120                             ct, ctinfo);
121 }
122
123 static unsigned int ipv4_confirm(unsigned int hooknum,
124                                  struct sk_buff *skb,
125                                  const struct net_device *in,
126                                  const struct net_device *out,
127                                  int (*okfn)(struct sk_buff *))
128 {
129         struct nf_conn *ct;
130         enum ip_conntrack_info ctinfo;
131
132         ct = nf_ct_get(skb, &ctinfo);
133         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
134                 goto out;
135
136         /* adjust seqs for loopback traffic only in outgoing direction */
137         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
138             !nf_is_loopback_packet(skb)) {
139                 typeof(nf_nat_seq_adjust_hook) seq_adjust;
140
141                 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
142                 if (!seq_adjust ||
143                     !seq_adjust(skb, ct, ctinfo, ip_hdrlen(skb))) {
144                         NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
145                         return NF_DROP;
146                 }
147         }
148 out:
149         /* We've seen it coming out the other side: confirm it */
150         return nf_conntrack_confirm(skb);
151 }
152
153 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
154                                       struct sk_buff *skb,
155                                       const struct net_device *in,
156                                       const struct net_device *out,
157                                       int (*okfn)(struct sk_buff *))
158 {
159         return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
160 }
161
162 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
163                                          struct sk_buff *skb,
164                                          const struct net_device *in,
165                                          const struct net_device *out,
166                                          int (*okfn)(struct sk_buff *))
167 {
168         /* root is playing with raw sockets. */
169         if (skb->len < sizeof(struct iphdr) ||
170             ip_hdrlen(skb) < sizeof(struct iphdr))
171                 return NF_ACCEPT;
172         return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
173 }
174
175 /* Connection tracking may drop packets, but never alters them, so
176    make it the first hook. */
177 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
178         {
179                 .hook           = ipv4_conntrack_in,
180                 .owner          = THIS_MODULE,
181                 .pf             = NFPROTO_IPV4,
182                 .hooknum        = NF_INET_PRE_ROUTING,
183                 .priority       = NF_IP_PRI_CONNTRACK,
184         },
185         {
186                 .hook           = ipv4_conntrack_local,
187                 .owner          = THIS_MODULE,
188                 .pf             = NFPROTO_IPV4,
189                 .hooknum        = NF_INET_LOCAL_OUT,
190                 .priority       = NF_IP_PRI_CONNTRACK,
191         },
192         {
193                 .hook           = ipv4_helper,
194                 .owner          = THIS_MODULE,
195                 .pf             = NFPROTO_IPV4,
196                 .hooknum        = NF_INET_POST_ROUTING,
197                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
198         },
199         {
200                 .hook           = ipv4_confirm,
201                 .owner          = THIS_MODULE,
202                 .pf             = NFPROTO_IPV4,
203                 .hooknum        = NF_INET_POST_ROUTING,
204                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
205         },
206         {
207                 .hook           = ipv4_helper,
208                 .owner          = THIS_MODULE,
209                 .pf             = NFPROTO_IPV4,
210                 .hooknum        = NF_INET_LOCAL_IN,
211                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
212         },
213         {
214                 .hook           = ipv4_confirm,
215                 .owner          = THIS_MODULE,
216                 .pf             = NFPROTO_IPV4,
217                 .hooknum        = NF_INET_LOCAL_IN,
218                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
219         },
220 };
221
222 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
223 static int log_invalid_proto_min = 0;
224 static int log_invalid_proto_max = 255;
225
226 static struct ctl_table ip_ct_sysctl_table[] = {
227         {
228                 .procname       = "ip_conntrack_max",
229                 .maxlen         = sizeof(int),
230                 .mode           = 0644,
231                 .proc_handler   = proc_dointvec,
232         },
233         {
234                 .procname       = "ip_conntrack_count",
235                 .maxlen         = sizeof(int),
236                 .mode           = 0444,
237                 .proc_handler   = proc_dointvec,
238         },
239         {
240                 .procname       = "ip_conntrack_buckets",
241                 .maxlen         = sizeof(unsigned int),
242                 .mode           = 0444,
243                 .proc_handler   = proc_dointvec,
244         },
245         {
246                 .procname       = "ip_conntrack_checksum",
247                 .maxlen         = sizeof(int),
248                 .mode           = 0644,
249                 .proc_handler   = proc_dointvec,
250         },
251         {
252                 .procname       = "ip_conntrack_log_invalid",
253                 .maxlen         = sizeof(unsigned int),
254                 .mode           = 0644,
255                 .proc_handler   = proc_dointvec_minmax,
256                 .extra1         = &log_invalid_proto_min,
257                 .extra2         = &log_invalid_proto_max,
258         },
259         { }
260 };
261 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
262
263 /* Fast function for those who don't want to parse /proc (and I don't
264    blame them). */
265 /* Reversing the socket's dst/src point of view gives us the reply
266    mapping. */
267 static int
268 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
269 {
270         const struct inet_sock *inet = inet_sk(sk);
271         const struct nf_conntrack_tuple_hash *h;
272         struct nf_conntrack_tuple tuple;
273
274         memset(&tuple, 0, sizeof(tuple));
275         tuple.src.u3.ip = inet->inet_rcv_saddr;
276         tuple.src.u.tcp.port = inet->inet_sport;
277         tuple.dst.u3.ip = inet->inet_daddr;
278         tuple.dst.u.tcp.port = inet->inet_dport;
279         tuple.src.l3num = PF_INET;
280         tuple.dst.protonum = sk->sk_protocol;
281
282         /* We only do TCP and SCTP at the moment: is there a better way? */
283         if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
284                 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
285                 return -ENOPROTOOPT;
286         }
287
288         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
289                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
290                          *len, sizeof(struct sockaddr_in));
291                 return -EINVAL;
292         }
293
294         h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
295         if (h) {
296                 struct sockaddr_in sin;
297                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
298
299                 sin.sin_family = AF_INET;
300                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
301                         .tuple.dst.u.tcp.port;
302                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
303                         .tuple.dst.u3.ip;
304                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
305
306                 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
307                          &sin.sin_addr.s_addr, ntohs(sin.sin_port));
308                 nf_ct_put(ct);
309                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
310                         return -EFAULT;
311                 else
312                         return 0;
313         }
314         pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
315                  &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
316                  &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
317         return -ENOENT;
318 }
319
320 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
321
322 #include <linux/netfilter/nfnetlink.h>
323 #include <linux/netfilter/nfnetlink_conntrack.h>
324
325 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
326                                 const struct nf_conntrack_tuple *tuple)
327 {
328         if (nla_put_be32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
329             nla_put_be32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
330                 goto nla_put_failure;
331         return 0;
332
333 nla_put_failure:
334         return -1;
335 }
336
337 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
338         [CTA_IP_V4_SRC] = { .type = NLA_U32 },
339         [CTA_IP_V4_DST] = { .type = NLA_U32 },
340 };
341
342 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
343                                 struct nf_conntrack_tuple *t)
344 {
345         if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
346                 return -EINVAL;
347
348         t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
349         t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
350
351         return 0;
352 }
353
354 static int ipv4_nlattr_tuple_size(void)
355 {
356         return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
357 }
358 #endif
359
360 static struct nf_sockopt_ops so_getorigdst = {
361         .pf             = PF_INET,
362         .get_optmin     = SO_ORIGINAL_DST,
363         .get_optmax     = SO_ORIGINAL_DST+1,
364         .get            = &getorigdst,
365         .owner          = THIS_MODULE,
366 };
367
368 static int ipv4_init_net(struct net *net)
369 {
370 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
371         struct nf_ip_net *in = &net->ct.nf_ct_proto;
372         in->ctl_table = kmemdup(ip_ct_sysctl_table,
373                                 sizeof(ip_ct_sysctl_table),
374                                 GFP_KERNEL);
375         if (!in->ctl_table)
376                 return -ENOMEM;
377
378         in->ctl_table[0].data = &nf_conntrack_max;
379         in->ctl_table[1].data = &net->ct.count;
380         in->ctl_table[2].data = &net->ct.htable_size;
381         in->ctl_table[3].data = &net->ct.sysctl_checksum;
382         in->ctl_table[4].data = &net->ct.sysctl_log_invalid;
383 #endif
384         return 0;
385 }
386
387 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
388         .l3proto         = PF_INET,
389         .name            = "ipv4",
390         .pkt_to_tuple    = ipv4_pkt_to_tuple,
391         .invert_tuple    = ipv4_invert_tuple,
392         .print_tuple     = ipv4_print_tuple,
393         .get_l4proto     = ipv4_get_l4proto,
394 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
395         .tuple_to_nlattr = ipv4_tuple_to_nlattr,
396         .nlattr_tuple_size = ipv4_nlattr_tuple_size,
397         .nlattr_to_tuple = ipv4_nlattr_to_tuple,
398         .nla_policy      = ipv4_nla_policy,
399 #endif
400 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
401         .ctl_table_path  = "net/ipv4/netfilter",
402 #endif
403         .init_net        = ipv4_init_net,
404         .me              = THIS_MODULE,
405 };
406
407 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
408                   &nf_conntrack_htable_size, 0600);
409
410 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
411 MODULE_ALIAS("ip_conntrack");
412 MODULE_LICENSE("GPL");
413
414 static int ipv4_net_init(struct net *net)
415 {
416         int ret = 0;
417
418         ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_tcp4);
419         if (ret < 0) {
420                 pr_err("nf_conntrack_tcp4: pernet registration failed\n");
421                 goto out_tcp;
422         }
423         ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_udp4);
424         if (ret < 0) {
425                 pr_err("nf_conntrack_udp4: pernet registration failed\n");
426                 goto out_udp;
427         }
428         ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_icmp);
429         if (ret < 0) {
430                 pr_err("nf_conntrack_icmp4: pernet registration failed\n");
431                 goto out_icmp;
432         }
433         ret = nf_ct_l3proto_pernet_register(net, &nf_conntrack_l3proto_ipv4);
434         if (ret < 0) {
435                 pr_err("nf_conntrack_ipv4: pernet registration failed\n");
436                 goto out_ipv4;
437         }
438         return 0;
439 out_ipv4:
440         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmp);
441 out_icmp:
442         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp4);
443 out_udp:
444         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp4);
445 out_tcp:
446         return ret;
447 }
448
449 static void ipv4_net_exit(struct net *net)
450 {
451         nf_ct_l3proto_pernet_unregister(net, &nf_conntrack_l3proto_ipv4);
452         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmp);
453         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp4);
454         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp4);
455 }
456
457 static struct pernet_operations ipv4_net_ops = {
458         .init = ipv4_net_init,
459         .exit = ipv4_net_exit,
460 };
461
462 static int __init nf_conntrack_l3proto_ipv4_init(void)
463 {
464         int ret = 0;
465
466         need_conntrack();
467         nf_defrag_ipv4_enable();
468
469         ret = nf_register_sockopt(&so_getorigdst);
470         if (ret < 0) {
471                 printk(KERN_ERR "Unable to register netfilter socket option\n");
472                 return ret;
473         }
474
475         ret = register_pernet_subsys(&ipv4_net_ops);
476         if (ret < 0) {
477                 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
478                 goto cleanup_sockopt;
479         }
480
481         ret = nf_register_hooks(ipv4_conntrack_ops,
482                                 ARRAY_SIZE(ipv4_conntrack_ops));
483         if (ret < 0) {
484                 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
485                 goto cleanup_pernet;
486         }
487
488         ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_tcp4);
489         if (ret < 0) {
490                 pr_err("nf_conntrack_ipv4: can't register tcp4 proto.\n");
491                 goto cleanup_hooks;
492         }
493
494         ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_udp4);
495         if (ret < 0) {
496                 pr_err("nf_conntrack_ipv4: can't register udp4 proto.\n");
497                 goto cleanup_tcp4;
498         }
499
500         ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_icmp);
501         if (ret < 0) {
502                 pr_err("nf_conntrack_ipv4: can't register icmpv4 proto.\n");
503                 goto cleanup_udp4;
504         }
505
506         ret = nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv4);
507         if (ret < 0) {
508                 pr_err("nf_conntrack_ipv4: can't register ipv4 proto.\n");
509                 goto cleanup_icmpv4;
510         }
511
512 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
513         ret = nf_conntrack_ipv4_compat_init();
514         if (ret < 0)
515                 goto cleanup_proto;
516 #endif
517         return ret;
518 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
519  cleanup_proto:
520         nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
521 #endif
522  cleanup_icmpv4:
523         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp);
524  cleanup_udp4:
525         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4);
526  cleanup_tcp4:
527         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
528  cleanup_hooks:
529         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
530  cleanup_pernet:
531         unregister_pernet_subsys(&ipv4_net_ops);
532  cleanup_sockopt:
533         nf_unregister_sockopt(&so_getorigdst);
534         return ret;
535 }
536
537 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
538 {
539         synchronize_net();
540 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
541         nf_conntrack_ipv4_compat_fini();
542 #endif
543         nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
544         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp);
545         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4);
546         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
547         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
548         unregister_pernet_subsys(&ipv4_net_ops);
549         nf_unregister_sockopt(&so_getorigdst);
550 }
551
552 module_init(nf_conntrack_l3proto_ipv4_init);
553 module_exit(nf_conntrack_l3proto_ipv4_fini);
554
555 void need_ipv4_conntrack(void)
556 {
557         return;
558 }
559 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);