?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
netfilter: nf_conntrack: fix hash resizing with namespaces
[linux-2.6/libata-dev.git] / net / ipv4 / netfilter / nf_conntrack_l3proto_ipv4.c
blobd1ea38a7c490befaaf23456ea8b774aa6ada6fd9
1
2 /* (C) 1999-2001 Paul `Rusty' Russell
3 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
8 */
9
10 #include <linux/types.h>
11 #include <linux/ip.h>
12 #include <linux/netfilter.h>
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/icmp.h>
16 #include <linux/sysctl.h>
17 #include <net/route.h>
18 #include <net/ip.h>
19
20 #include <linux/netfilter_ipv4.h>
21 #include <net/netfilter/nf_conntrack.h>
22 #include <net/netfilter/nf_conntrack_helper.h>
23 #include <net/netfilter/nf_conntrack_l4proto.h>
24 #include <net/netfilter/nf_conntrack_l3proto.h>
25 #include <net/netfilter/nf_conntrack_core.h>
26 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
27 #include <net/netfilter/nf_nat_helper.h>
28 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
29 #include <net/netfilter/nf_log.h>
30
31 int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
32 struct nf_conn *ct,
33 enum ip_conntrack_info ctinfo);
34 EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
35
36 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
37 struct nf_conntrack_tuple *tuple)
38 {
39 const __be32 *ap;
40 __be32 _addrs[2];
41 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
42 sizeof(u_int32_t) * 2, _addrs);
43 if (ap == NULL)
44 return false;
45
46 tuple->src.u3.ip = ap[0];
47 tuple->dst.u3.ip = ap[1];
48
49 return true;
50 }
51
52 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
53 const struct nf_conntrack_tuple *orig)
54 {
55 tuple->src.u3.ip = orig->dst.u3.ip;
56 tuple->dst.u3.ip = orig->src.u3.ip;
57
58 return true;
59 }
60
61 static int ipv4_print_tuple(struct seq_file *s,
62 const struct nf_conntrack_tuple *tuple)
63 {
64 return seq_printf(s, "src=%pI4 dst=%pI4 ",
65 &tuple->src.u3.ip, &tuple->dst.u3.ip);
66 }
67
68 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
69 unsigned int *dataoff, u_int8_t *protonum)
70 {
71 const struct iphdr *iph;
72 struct iphdr _iph;
73
74 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
75 if (iph == NULL)
76 return -NF_DROP;
77
78 /* Conntrack defragments packets, we might still see fragments
79 * inside ICMP packets though. */
80 if (iph->frag_off & htons(IP_OFFSET))
81 return -NF_DROP;
82
83 *dataoff = nhoff + (iph->ihl << 2);
84 *protonum = iph->protocol;
85
86 return NF_ACCEPT;
87 }
88
89 static unsigned int ipv4_confirm(unsigned int hooknum,
90 struct sk_buff *skb,
91 const struct net_device *in,
92 const struct net_device *out,
93 int (*okfn)(struct sk_buff *))
94 {
95 struct nf_conn *ct;
96 enum ip_conntrack_info ctinfo;
97 const struct nf_conn_help *help;
98 const struct nf_conntrack_helper *helper;
99 unsigned int ret;
100
101 /* This is where we call the helper: as the packet goes out. */
102 ct = nf_ct_get(skb, &ctinfo);
103 if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
104 goto out;
105
106 help = nfct_help(ct);
107 if (!help)
108 goto out;
109
110 /* rcu_read_lock()ed by nf_hook_slow */
111 helper = rcu_dereference(help->helper);
112 if (!helper)
113 goto out;
114
115 ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
116 ct, ctinfo);
117 if (ret != NF_ACCEPT) {
118 nf_log_packet(NFPROTO_IPV4, hooknum, skb, in, out, NULL,
119 "nf_ct_%s: dropping packet", helper->name);
120 return ret;
121 }
122
123 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
124 typeof(nf_nat_seq_adjust_hook) seq_adjust;
125
126 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
127 if (!seq_adjust || !seq_adjust(skb, ct, ctinfo)) {
128 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
129 return NF_DROP;
130 }
131 }
132 out:
133 /* We've seen it coming out the other side: confirm it */
134 return nf_conntrack_confirm(skb);
135 }
136
137 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
138 struct sk_buff *skb,
139 const struct net_device *in,
140 const struct net_device *out,
141 int (*okfn)(struct sk_buff *))
142 {
143 return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
144 }
145
146 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
147 struct sk_buff *skb,
148 const struct net_device *in,
149 const struct net_device *out,
150 int (*okfn)(struct sk_buff *))
151 {
152 /* root is playing with raw sockets. */
153 if (skb->len < sizeof(struct iphdr) ||
154 ip_hdrlen(skb) < sizeof(struct iphdr))
155 return NF_ACCEPT;
156 return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
157 }
158
159 /* Connection tracking may drop packets, but never alters them, so
160 make it the first hook. */
161 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
162 {
163 .hook = ipv4_conntrack_in,
164 .owner = THIS_MODULE,
165 .pf = NFPROTO_IPV4,
166 .hooknum = NF_INET_PRE_ROUTING,
167 .priority = NF_IP_PRI_CONNTRACK,
168 },
169 {
170 .hook = ipv4_conntrack_local,
171 .owner = THIS_MODULE,
172 .pf = NFPROTO_IPV4,
173 .hooknum = NF_INET_LOCAL_OUT,
174 .priority = NF_IP_PRI_CONNTRACK,
175 },
176 {
177 .hook = ipv4_confirm,
178 .owner = THIS_MODULE,
179 .pf = NFPROTO_IPV4,
180 .hooknum = NF_INET_POST_ROUTING,
181 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
182 },
183 {
184 .hook = ipv4_confirm,
185 .owner = THIS_MODULE,
186 .pf = NFPROTO_IPV4,
187 .hooknum = NF_INET_LOCAL_IN,
188 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
189 },
190 };
191
192 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
193 static int log_invalid_proto_min = 0;
194 static int log_invalid_proto_max = 255;
195
196 static ctl_table ip_ct_sysctl_table[] = {
197 {
198 .procname = "ip_conntrack_max",
199 .data = &nf_conntrack_max,
200 .maxlen = sizeof(int),
201 .mode = 0644,
202 .proc_handler = proc_dointvec,
203 },
204 {
205 .procname = "ip_conntrack_count",
206 .data = &init_net.ct.count,
207 .maxlen = sizeof(int),
208 .mode = 0444,
209 .proc_handler = proc_dointvec,
210 },
211 {
212 .procname = "ip_conntrack_buckets",
213 .data = &init_net.ct.htable_size,
214 .maxlen = sizeof(unsigned int),
215 .mode = 0444,
216 .proc_handler = proc_dointvec,
217 },
218 {
219 .procname = "ip_conntrack_checksum",
220 .data = &init_net.ct.sysctl_checksum,
221 .maxlen = sizeof(int),
222 .mode = 0644,
223 .proc_handler = proc_dointvec,
224 },
225 {
226 .procname = "ip_conntrack_log_invalid",
227 .data = &init_net.ct.sysctl_log_invalid,
228 .maxlen = sizeof(unsigned int),
229 .mode = 0644,
230 .proc_handler = proc_dointvec_minmax,
231 .extra1 = &log_invalid_proto_min,
232 .extra2 = &log_invalid_proto_max,
233 },
234 { }
235 };
236 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
237
238 /* Fast function for those who don't want to parse /proc (and I don't
239 blame them). */
240 /* Reversing the socket's dst/src point of view gives us the reply
241 mapping. */
242 static int
243 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
244 {
245 const struct inet_sock *inet = inet_sk(sk);
246 const struct nf_conntrack_tuple_hash *h;
247 struct nf_conntrack_tuple tuple;
248
249 memset(&tuple, 0, sizeof(tuple));
250 tuple.src.u3.ip = inet->inet_rcv_saddr;
251 tuple.src.u.tcp.port = inet->inet_sport;
252 tuple.dst.u3.ip = inet->inet_daddr;
253 tuple.dst.u.tcp.port = inet->inet_dport;
254 tuple.src.l3num = PF_INET;
255 tuple.dst.protonum = sk->sk_protocol;
256
257 /* We only do TCP and SCTP at the moment: is there a better way? */
258 if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
259 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
260 return -ENOPROTOOPT;
261 }
262
263 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
264 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
265 *len, sizeof(struct sockaddr_in));
266 return -EINVAL;
267 }
268
269 h = nf_conntrack_find_get(sock_net(sk), &tuple);
270 if (h) {
271 struct sockaddr_in sin;
272 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
273
274 sin.sin_family = AF_INET;
275 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
276 .tuple.dst.u.tcp.port;
277 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
278 .tuple.dst.u3.ip;
279 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
280
281 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
282 &sin.sin_addr.s_addr, ntohs(sin.sin_port));
283 nf_ct_put(ct);
284 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
285 return -EFAULT;
286 else
287 return 0;
288 }
289 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
290 &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
291 &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
292 return -ENOENT;
293 }
294
295 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
296
297 #include <linux/netfilter/nfnetlink.h>
298 #include <linux/netfilter/nfnetlink_conntrack.h>
299
300 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
301 const struct nf_conntrack_tuple *tuple)
302 {
303 NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
304 NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
305 return 0;
306
307 nla_put_failure:
308 return -1;
309 }
310
311 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
312 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
313 [CTA_IP_V4_DST] = { .type = NLA_U32 },
314 };
315
316 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
317 struct nf_conntrack_tuple *t)
318 {
319 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
320 return -EINVAL;
321
322 t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
323 t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
324
325 return 0;
326 }
327
328 static int ipv4_nlattr_tuple_size(void)
329 {
330 return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
331 }
332 #endif
333
334 static struct nf_sockopt_ops so_getorigdst = {
335 .pf = PF_INET,
336 .get_optmin = SO_ORIGINAL_DST,
337 .get_optmax = SO_ORIGINAL_DST+1,
338 .get = &getorigdst,
339 .owner = THIS_MODULE,
340 };
341
342 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
343 .l3proto = PF_INET,
344 .name = "ipv4",
345 .pkt_to_tuple = ipv4_pkt_to_tuple,
346 .invert_tuple = ipv4_invert_tuple,
347 .print_tuple = ipv4_print_tuple,
348 .get_l4proto = ipv4_get_l4proto,
349 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
350 .tuple_to_nlattr = ipv4_tuple_to_nlattr,
351 .nlattr_tuple_size = ipv4_nlattr_tuple_size,
352 .nlattr_to_tuple = ipv4_nlattr_to_tuple,
353 .nla_policy = ipv4_nla_policy,
354 #endif
355 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
356 .ctl_table_path = nf_net_ipv4_netfilter_sysctl_path,
357 .ctl_table = ip_ct_sysctl_table,
358 #endif
359 .me = THIS_MODULE,
360 };
361
362 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
363 &nf_conntrack_htable_size, 0600);
364
365 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
366 MODULE_ALIAS("ip_conntrack");
367 MODULE_LICENSE("GPL");
368
369 static int __init nf_conntrack_l3proto_ipv4_init(void)
370 {
371 int ret = 0;
372
373 need_conntrack();
374 nf_defrag_ipv4_enable();
375
376 ret = nf_register_sockopt(&so_getorigdst);
377 if (ret < 0) {
378 printk(KERN_ERR "Unable to register netfilter socket option\n");
379 return ret;
380 }
381
382 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
383 if (ret < 0) {
384 printk("nf_conntrack_ipv4: can't register tcp.\n");
385 goto cleanup_sockopt;
386 }
387
388 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
389 if (ret < 0) {
390 printk("nf_conntrack_ipv4: can't register udp.\n");
391 goto cleanup_tcp;
392 }
393
394 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
395 if (ret < 0) {
396 printk("nf_conntrack_ipv4: can't register icmp.\n");
397 goto cleanup_udp;
398 }
399
400 ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
401 if (ret < 0) {
402 printk("nf_conntrack_ipv4: can't register ipv4\n");
403 goto cleanup_icmp;
404 }
405
406 ret = nf_register_hooks(ipv4_conntrack_ops,
407 ARRAY_SIZE(ipv4_conntrack_ops));
408 if (ret < 0) {
409 printk("nf_conntrack_ipv4: can't register hooks.\n");
410 goto cleanup_ipv4;
411 }
412 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
413 ret = nf_conntrack_ipv4_compat_init();
414 if (ret < 0)
415 goto cleanup_hooks;
416 #endif
417 return ret;
418 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
419 cleanup_hooks:
420 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
421 #endif
422 cleanup_ipv4:
423 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
424 cleanup_icmp:
425 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
426 cleanup_udp:
427 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
428 cleanup_tcp:
429 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
430 cleanup_sockopt:
431 nf_unregister_sockopt(&so_getorigdst);
432 return ret;
433 }
434
435 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
436 {
437 synchronize_net();
438 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
439 nf_conntrack_ipv4_compat_fini();
440 #endif
441 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
442 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
443 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
444 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
445 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
446 nf_unregister_sockopt(&so_getorigdst);
447 }
448
449 module_init(nf_conntrack_l3proto_ipv4_init);
450 module_exit(nf_conntrack_l3proto_ipv4_fini);
451
452 void need_ipv4_conntrack(void)
453 {
454 return;
455 }
456 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);
Linux 2.6 libata-dev (mirror)