netns xfrm: fix "ip xfrm state|policy count" misreport
[safe/jmp/linux-2.6] / net / sched / act_ipt.c
1 /*
2  * net/sched/ipt.c      iptables target interface
3  *
4  *TODO: Add other tables. For now we only support the ipv4 table targets
5  *
6  *              This program is free software; you can redistribute it and/or
7  *              modify it under the terms of the GNU General Public License
8  *              as published by the Free Software Foundation; either version
9  *              2 of the License, or (at your option) any later version.
10  *
11  * Copyright:   Jamal Hadi Salim (2002-4)
12  */
13
14 #include <linux/types.h>
15 #include <linux/kernel.h>
16 #include <linux/string.h>
17 #include <linux/errno.h>
18 #include <linux/skbuff.h>
19 #include <linux/rtnetlink.h>
20 #include <linux/module.h>
21 #include <linux/init.h>
22 #include <net/netlink.h>
23 #include <net/pkt_sched.h>
24 #include <linux/tc_act/tc_ipt.h>
25 #include <net/tc_act/tc_ipt.h>
26
27 #include <linux/netfilter_ipv4/ip_tables.h>
28
29
30 #define IPT_TAB_MASK     15
31 static struct tcf_common *tcf_ipt_ht[IPT_TAB_MASK + 1];
32 static u32 ipt_idx_gen;
33 static DEFINE_RWLOCK(ipt_lock);
34
35 static struct tcf_hashinfo ipt_hash_info = {
36         .htab   =       tcf_ipt_ht,
37         .hmask  =       IPT_TAB_MASK,
38         .lock   =       &ipt_lock,
39 };
40
41 static int ipt_init_target(struct ipt_entry_target *t, char *table, unsigned int hook)
42 {
43         struct xt_tgchk_param par;
44         struct xt_target *target;
45         int ret = 0;
46
47         target = xt_request_find_target(AF_INET, t->u.user.name,
48                                         t->u.user.revision);
49         if (!target)
50                 return -ENOENT;
51
52         t->u.kernel.target = target;
53         par.table     = table;
54         par.entryinfo = NULL;
55         par.target    = target;
56         par.targinfo  = t->data;
57         par.hook_mask = hook;
58         par.family    = NFPROTO_IPV4;
59
60         ret = xt_check_target(&par, t->u.target_size - sizeof(*t), 0, false);
61         if (ret < 0) {
62                 module_put(t->u.kernel.target->me);
63                 return ret;
64         }
65         return 0;
66 }
67
68 static void ipt_destroy_target(struct ipt_entry_target *t)
69 {
70         struct xt_tgdtor_param par = {
71                 .target   = t->u.kernel.target,
72                 .targinfo = t->data,
73         };
74         if (par.target->destroy != NULL)
75                 par.target->destroy(&par);
76         module_put(par.target->me);
77 }
78
79 static int tcf_ipt_release(struct tcf_ipt *ipt, int bind)
80 {
81         int ret = 0;
82         if (ipt) {
83                 if (bind)
84                         ipt->tcf_bindcnt--;
85                 ipt->tcf_refcnt--;
86                 if (ipt->tcf_bindcnt <= 0 && ipt->tcf_refcnt <= 0) {
87                         ipt_destroy_target(ipt->tcfi_t);
88                         kfree(ipt->tcfi_tname);
89                         kfree(ipt->tcfi_t);
90                         tcf_hash_destroy(&ipt->common, &ipt_hash_info);
91                         ret = ACT_P_DELETED;
92                 }
93         }
94         return ret;
95 }
96
97 static const struct nla_policy ipt_policy[TCA_IPT_MAX + 1] = {
98         [TCA_IPT_TABLE] = { .type = NLA_STRING, .len = IFNAMSIZ },
99         [TCA_IPT_HOOK]  = { .type = NLA_U32 },
100         [TCA_IPT_INDEX] = { .type = NLA_U32 },
101         [TCA_IPT_TARG]  = { .len = sizeof(struct ipt_entry_target) },
102 };
103
104 static int tcf_ipt_init(struct nlattr *nla, struct nlattr *est,
105                         struct tc_action *a, int ovr, int bind)
106 {
107         struct nlattr *tb[TCA_IPT_MAX + 1];
108         struct tcf_ipt *ipt;
109         struct tcf_common *pc;
110         struct ipt_entry_target *td, *t;
111         char *tname;
112         int ret = 0, err;
113         u32 hook = 0;
114         u32 index = 0;
115
116         if (nla == NULL)
117                 return -EINVAL;
118
119         err = nla_parse_nested(tb, TCA_IPT_MAX, nla, ipt_policy);
120         if (err < 0)
121                 return err;
122
123         if (tb[TCA_IPT_HOOK] == NULL)
124                 return -EINVAL;
125         if (tb[TCA_IPT_TARG] == NULL)
126                 return -EINVAL;
127
128         td = (struct ipt_entry_target *)nla_data(tb[TCA_IPT_TARG]);
129         if (nla_len(tb[TCA_IPT_TARG]) < td->u.target_size)
130                 return -EINVAL;
131
132         if (tb[TCA_IPT_INDEX] != NULL)
133                 index = nla_get_u32(tb[TCA_IPT_INDEX]);
134
135         pc = tcf_hash_check(index, a, bind, &ipt_hash_info);
136         if (!pc) {
137                 pc = tcf_hash_create(index, est, a, sizeof(*ipt), bind,
138                                      &ipt_idx_gen, &ipt_hash_info);
139                 if (IS_ERR(pc))
140                     return PTR_ERR(pc);
141                 ret = ACT_P_CREATED;
142         } else {
143                 if (!ovr) {
144                         tcf_ipt_release(to_ipt(pc), bind);
145                         return -EEXIST;
146                 }
147         }
148         ipt = to_ipt(pc);
149
150         hook = nla_get_u32(tb[TCA_IPT_HOOK]);
151
152         err = -ENOMEM;
153         tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
154         if (unlikely(!tname))
155                 goto err1;
156         if (tb[TCA_IPT_TABLE] == NULL ||
157             nla_strlcpy(tname, tb[TCA_IPT_TABLE], IFNAMSIZ) >= IFNAMSIZ)
158                 strcpy(tname, "mangle");
159
160         t = kmemdup(td, td->u.target_size, GFP_KERNEL);
161         if (unlikely(!t))
162                 goto err2;
163
164         if ((err = ipt_init_target(t, tname, hook)) < 0)
165                 goto err3;
166
167         spin_lock_bh(&ipt->tcf_lock);
168         if (ret != ACT_P_CREATED) {
169                 ipt_destroy_target(ipt->tcfi_t);
170                 kfree(ipt->tcfi_tname);
171                 kfree(ipt->tcfi_t);
172         }
173         ipt->tcfi_tname = tname;
174         ipt->tcfi_t     = t;
175         ipt->tcfi_hook  = hook;
176         spin_unlock_bh(&ipt->tcf_lock);
177         if (ret == ACT_P_CREATED)
178                 tcf_hash_insert(pc, &ipt_hash_info);
179         return ret;
180
181 err3:
182         kfree(t);
183 err2:
184         kfree(tname);
185 err1:
186         kfree(pc);
187         return err;
188 }
189
190 static int tcf_ipt_cleanup(struct tc_action *a, int bind)
191 {
192         struct tcf_ipt *ipt = a->priv;
193         return tcf_ipt_release(ipt, bind);
194 }
195
196 static int tcf_ipt(struct sk_buff *skb, struct tc_action *a,
197                    struct tcf_result *res)
198 {
199         int ret = 0, result = 0;
200         struct tcf_ipt *ipt = a->priv;
201         struct xt_target_param par;
202
203         if (skb_cloned(skb)) {
204                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
205                         return TC_ACT_UNSPEC;
206         }
207
208         spin_lock(&ipt->tcf_lock);
209
210         ipt->tcf_tm.lastuse = jiffies;
211         ipt->tcf_bstats.bytes += qdisc_pkt_len(skb);
212         ipt->tcf_bstats.packets++;
213
214         /* yes, we have to worry about both in and out dev
215          worry later - danger - this API seems to have changed
216          from earlier kernels */
217         par.in       = skb->dev;
218         par.out      = NULL;
219         par.hooknum  = ipt->tcfi_hook;
220         par.target   = ipt->tcfi_t->u.kernel.target;
221         par.targinfo = ipt->tcfi_t->data;
222         ret = par.target->target(skb, &par);
223
224         switch (ret) {
225         case NF_ACCEPT:
226                 result = TC_ACT_OK;
227                 break;
228         case NF_DROP:
229                 result = TC_ACT_SHOT;
230                 ipt->tcf_qstats.drops++;
231                 break;
232         case IPT_CONTINUE:
233                 result = TC_ACT_PIPE;
234                 break;
235         default:
236                 if (net_ratelimit())
237                         printk("Bogus netfilter code %d assume ACCEPT\n", ret);
238                 result = TC_POLICE_OK;
239                 break;
240         }
241         spin_unlock(&ipt->tcf_lock);
242         return result;
243
244 }
245
246 static int tcf_ipt_dump(struct sk_buff *skb, struct tc_action *a, int bind, int ref)
247 {
248         unsigned char *b = skb_tail_pointer(skb);
249         struct tcf_ipt *ipt = a->priv;
250         struct ipt_entry_target *t;
251         struct tcf_t tm;
252         struct tc_cnt c;
253
254         /* for simple targets kernel size == user size
255         ** user name = target name
256         ** for foolproof you need to not assume this
257         */
258
259         t = kmemdup(ipt->tcfi_t, ipt->tcfi_t->u.user.target_size, GFP_ATOMIC);
260         if (unlikely(!t))
261                 goto nla_put_failure;
262
263         c.bindcnt = ipt->tcf_bindcnt - bind;
264         c.refcnt = ipt->tcf_refcnt - ref;
265         strcpy(t->u.user.name, ipt->tcfi_t->u.kernel.target->name);
266
267         NLA_PUT(skb, TCA_IPT_TARG, ipt->tcfi_t->u.user.target_size, t);
268         NLA_PUT_U32(skb, TCA_IPT_INDEX, ipt->tcf_index);
269         NLA_PUT_U32(skb, TCA_IPT_HOOK, ipt->tcfi_hook);
270         NLA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
271         NLA_PUT_STRING(skb, TCA_IPT_TABLE, ipt->tcfi_tname);
272         tm.install = jiffies_to_clock_t(jiffies - ipt->tcf_tm.install);
273         tm.lastuse = jiffies_to_clock_t(jiffies - ipt->tcf_tm.lastuse);
274         tm.expires = jiffies_to_clock_t(ipt->tcf_tm.expires);
275         NLA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
276         kfree(t);
277         return skb->len;
278
279 nla_put_failure:
280         nlmsg_trim(skb, b);
281         kfree(t);
282         return -1;
283 }
284
285 static struct tc_action_ops act_ipt_ops = {
286         .kind           =       "ipt",
287         .hinfo          =       &ipt_hash_info,
288         .type           =       TCA_ACT_IPT,
289         .capab          =       TCA_CAP_NONE,
290         .owner          =       THIS_MODULE,
291         .act            =       tcf_ipt,
292         .dump           =       tcf_ipt_dump,
293         .cleanup        =       tcf_ipt_cleanup,
294         .lookup         =       tcf_hash_search,
295         .init           =       tcf_ipt_init,
296         .walk           =       tcf_generic_walker
297 };
298
299 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
300 MODULE_DESCRIPTION("Iptables target actions");
301 MODULE_LICENSE("GPL");
302
303 static int __init ipt_init_module(void)
304 {
305         return tcf_register_action(&act_ipt_ops);
306 }
307
308 static void __exit ipt_cleanup_module(void)
309 {
310         tcf_unregister_action(&act_ipt_ops);
311 }
312
313 module_init(ipt_init_module);
314 module_exit(ipt_cleanup_module);