Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/bnx2-2.6
[sfrench/cifs-2.6.git] / 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 <asm/uaccess.h>
15 #include <asm/system.h>
16 #include <asm/bitops.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/sched.h>
20 #include <linux/string.h>
21 #include <linux/mm.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/in.h>
25 #include <linux/errno.h>
26 #include <linux/interrupt.h>
27 #include <linux/netdevice.h>
28 #include <linux/skbuff.h>
29 #include <linux/rtnetlink.h>
30 #include <linux/module.h>
31 #include <linux/init.h>
32 #include <linux/proc_fs.h>
33 #include <linux/kmod.h>
34 #include <net/sock.h>
35 #include <net/pkt_sched.h>
36 #include <linux/tc_act/tc_ipt.h>
37 #include <net/tc_act/tc_ipt.h>
38
39 #include <linux/netfilter_ipv4/ip_tables.h>
40
41
42 #define IPT_TAB_MASK     15
43 static struct tcf_common *tcf_ipt_ht[IPT_TAB_MASK + 1];
44 static u32 ipt_idx_gen;
45 static DEFINE_RWLOCK(ipt_lock);
46
47 static struct tcf_hashinfo ipt_hash_info = {
48         .htab   =       tcf_ipt_ht,
49         .hmask  =       IPT_TAB_MASK,
50         .lock   =       &ipt_lock,
51 };
52
53 static int ipt_init_target(struct ipt_entry_target *t, char *table, unsigned int hook)
54 {
55         struct ipt_target *target;
56         int ret = 0;
57
58         target = xt_request_find_target(AF_INET, t->u.user.name,
59                                         t->u.user.revision);
60         if (!target)
61                 return -ENOENT;
62
63         t->u.kernel.target = target;
64
65         ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
66                               table, hook, 0, 0);
67         if (ret) {
68                 module_put(t->u.kernel.target->me);
69                 return ret;
70         }
71         if (t->u.kernel.target->checkentry
72             && !t->u.kernel.target->checkentry(table, NULL,
73                                                t->u.kernel.target, t->data,
74                                                hook)) {
75                 module_put(t->u.kernel.target->me);
76                 ret = -EINVAL;
77         }
78
79         return ret;
80 }
81
82 static void ipt_destroy_target(struct ipt_entry_target *t)
83 {
84         if (t->u.kernel.target->destroy)
85                 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
86         module_put(t->u.kernel.target->me);
87 }
88
89 static int tcf_ipt_release(struct tcf_ipt *ipt, int bind)
90 {
91         int ret = 0;
92         if (ipt) {
93                 if (bind)
94                         ipt->tcf_bindcnt--;
95                 ipt->tcf_refcnt--;
96                 if (ipt->tcf_bindcnt <= 0 && ipt->tcf_refcnt <= 0) {
97                         ipt_destroy_target(ipt->tcfi_t);
98                         kfree(ipt->tcfi_tname);
99                         kfree(ipt->tcfi_t);
100                         tcf_hash_destroy(&ipt->common, &ipt_hash_info);
101                         ret = ACT_P_DELETED;
102                 }
103         }
104         return ret;
105 }
106
107 static int tcf_ipt_init(struct rtattr *rta, struct rtattr *est,
108                         struct tc_action *a, int ovr, int bind)
109 {
110         struct rtattr *tb[TCA_IPT_MAX];
111         struct tcf_ipt *ipt;
112         struct tcf_common *pc;
113         struct ipt_entry_target *td, *t;
114         char *tname;
115         int ret = 0, err;
116         u32 hook = 0;
117         u32 index = 0;
118
119         if (rta == NULL || rtattr_parse_nested(tb, TCA_IPT_MAX, rta) < 0)
120                 return -EINVAL;
121
122         if (tb[TCA_IPT_HOOK-1] == NULL ||
123             RTA_PAYLOAD(tb[TCA_IPT_HOOK-1]) < sizeof(u32))
124                 return -EINVAL;
125         if (tb[TCA_IPT_TARG-1] == NULL ||
126             RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < sizeof(*t))
127                 return -EINVAL;
128         td = (struct ipt_entry_target *)RTA_DATA(tb[TCA_IPT_TARG-1]);
129         if (RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < td->u.target_size)
130                 return -EINVAL;
131
132         if (tb[TCA_IPT_INDEX-1] != NULL &&
133             RTA_PAYLOAD(tb[TCA_IPT_INDEX-1]) >= sizeof(u32))
134                 index = *(u32 *)RTA_DATA(tb[TCA_IPT_INDEX-1]);
135
136         pc = tcf_hash_check(index, a, bind, &ipt_hash_info);
137         if (!pc) {
138                 pc = tcf_hash_create(index, est, a, sizeof(*ipt), bind,
139                                      &ipt_idx_gen, &ipt_hash_info);
140                 if (unlikely(!pc))
141                         return -ENOMEM;
142                 ret = ACT_P_CREATED;
143         } else {
144                 if (!ovr) {
145                         tcf_ipt_release(to_ipt(pc), bind);
146                         return -EEXIST;
147                 }
148         }
149         ipt = to_ipt(pc);
150
151         hook = *(u32 *)RTA_DATA(tb[TCA_IPT_HOOK-1]);
152
153         err = -ENOMEM;
154         tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
155         if (unlikely(!tname))
156                 goto err1;
157         if (tb[TCA_IPT_TABLE - 1] == NULL ||
158             rtattr_strlcpy(tname, tb[TCA_IPT_TABLE-1], IFNAMSIZ) >= IFNAMSIZ)
159                 strcpy(tname, "mangle");
160
161         t = kmemdup(td, td->u.target_size, GFP_KERNEL);
162         if (unlikely(!t))
163                 goto err2;
164
165         if ((err = ipt_init_target(t, tname, hook)) < 0)
166                 goto err3;
167
168         spin_lock_bh(&ipt->tcf_lock);
169         if (ret != ACT_P_CREATED) {
170                 ipt_destroy_target(ipt->tcfi_t);
171                 kfree(ipt->tcfi_tname);
172                 kfree(ipt->tcfi_t);
173         }
174         ipt->tcfi_tname = tname;
175         ipt->tcfi_t     = t;
176         ipt->tcfi_hook  = hook;
177         spin_unlock_bh(&ipt->tcf_lock);
178         if (ret == ACT_P_CREATED)
179                 tcf_hash_insert(pc, &ipt_hash_info);
180         return ret;
181
182 err3:
183         kfree(t);
184 err2:
185         kfree(tname);
186 err1:
187         kfree(pc);
188         return err;
189 }
190
191 static int tcf_ipt_cleanup(struct tc_action *a, int bind)
192 {
193         struct tcf_ipt *ipt = a->priv;
194         return tcf_ipt_release(ipt, bind);
195 }
196
197 static int tcf_ipt(struct sk_buff *skb, struct tc_action *a,
198                    struct tcf_result *res)
199 {
200         int ret = 0, result = 0;
201         struct tcf_ipt *ipt = a->priv;
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 += skb->len;
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
218         /* iptables targets take a double skb pointer in case the skb
219          * needs to be replaced. We don't own the skb, so this must not
220          * happen. The pskb_expand_head above should make sure of this */
221         ret = ipt->tcfi_t->u.kernel.target->target(&skb, skb->dev, NULL,
222                                                    ipt->tcfi_hook,
223                                                    ipt->tcfi_t->u.kernel.target,
224                                                    ipt->tcfi_t->data);
225         switch (ret) {
226         case NF_ACCEPT:
227                 result = TC_ACT_OK;
228                 break;
229         case NF_DROP:
230                 result = TC_ACT_SHOT;
231                 ipt->tcf_qstats.drops++;
232                 break;
233         case IPT_CONTINUE:
234                 result = TC_ACT_PIPE;
235                 break;
236         default:
237                 if (net_ratelimit())
238                         printk("Bogus netfilter code %d assume ACCEPT\n", ret);
239                 result = TC_POLICE_OK;
240                 break;
241         }
242         spin_unlock(&ipt->tcf_lock);
243         return result;
244
245 }
246
247 static int tcf_ipt_dump(struct sk_buff *skb, struct tc_action *a, int bind, int ref)
248 {
249         unsigned char *b = skb->tail;
250         struct tcf_ipt *ipt = a->priv;
251         struct ipt_entry_target *t;
252         struct tcf_t tm;
253         struct tc_cnt c;
254
255         /* for simple targets kernel size == user size
256         ** user name = target name
257         ** for foolproof you need to not assume this
258         */
259
260         t = kmemdup(ipt->tcfi_t, ipt->tcfi_t->u.user.target_size, GFP_ATOMIC);
261         if (unlikely(!t))
262                 goto rtattr_failure;
263
264         c.bindcnt = ipt->tcf_bindcnt - bind;
265         c.refcnt = ipt->tcf_refcnt - ref;
266         strcpy(t->u.user.name, ipt->tcfi_t->u.kernel.target->name);
267
268         RTA_PUT(skb, TCA_IPT_TARG, ipt->tcfi_t->u.user.target_size, t);
269         RTA_PUT(skb, TCA_IPT_INDEX, 4, &ipt->tcf_index);
270         RTA_PUT(skb, TCA_IPT_HOOK, 4, &ipt->tcfi_hook);
271         RTA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
272         RTA_PUT(skb, TCA_IPT_TABLE, IFNAMSIZ, ipt->tcfi_tname);
273         tm.install = jiffies_to_clock_t(jiffies - ipt->tcf_tm.install);
274         tm.lastuse = jiffies_to_clock_t(jiffies - ipt->tcf_tm.lastuse);
275         tm.expires = jiffies_to_clock_t(ipt->tcf_tm.expires);
276         RTA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
277         kfree(t);
278         return skb->len;
279
280 rtattr_failure:
281         skb_trim(skb, b - skb->data);
282         kfree(t);
283         return -1;
284 }
285
286 static struct tc_action_ops act_ipt_ops = {
287         .kind           =       "ipt",
288         .hinfo          =       &ipt_hash_info,
289         .type           =       TCA_ACT_IPT,
290         .capab          =       TCA_CAP_NONE,
291         .owner          =       THIS_MODULE,
292         .act            =       tcf_ipt,
293         .dump           =       tcf_ipt_dump,
294         .cleanup        =       tcf_ipt_cleanup,
295         .lookup         =       tcf_hash_search,
296         .init           =       tcf_ipt_init,
297         .walk           =       tcf_generic_walker
298 };
299
300 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
301 MODULE_DESCRIPTION("Iptables target actions");
302 MODULE_LICENSE("GPL");
303
304 static int __init ipt_init_module(void)
305 {
306         return tcf_register_action(&act_ipt_ops);
307 }
308
309 static void __exit ipt_cleanup_module(void)
310 {
311         tcf_unregister_action(&act_ipt_ops);
312 }
313
314 module_init(ipt_init_module);
315 module_exit(ipt_cleanup_module);