Merge branch 'devel' of master.kernel.org:/home/rmk/linux-2.6-serial
[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 /* use generic hash table */
42 #define MY_TAB_SIZE     16
43 #define MY_TAB_MASK     15
44
45 static u32 idx_gen;
46 static struct tcf_ipt *tcf_ipt_ht[MY_TAB_SIZE];
47 /* ipt hash table lock */
48 static DEFINE_RWLOCK(ipt_lock);
49
50 /* ovewrride the defaults */
51 #define tcf_st          tcf_ipt
52 #define tcf_t_lock      ipt_lock
53 #define tcf_ht          tcf_ipt_ht
54
55 #define CONFIG_NET_ACT_INIT
56 #include <net/pkt_act.h>
57
58 static int
59 ipt_init_target(struct ipt_entry_target *t, char *table, unsigned int hook)
60 {
61         struct ipt_target *target;
62         int ret = 0;
63
64         target = xt_find_target(AF_INET, t->u.user.name, t->u.user.revision);
65         if (!target)
66                 return -ENOENT;
67
68         DPRINTK("ipt_init_target: found %s\n", target->name);
69         t->u.kernel.target = target;
70
71         ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
72                               table, hook, 0, 0);
73         if (ret)
74                 return ret;
75
76         if (t->u.kernel.target->checkentry
77             && !t->u.kernel.target->checkentry(table, NULL,
78                                                t->u.kernel.target, t->data,
79                                                t->u.target_size - sizeof(*t),
80                                                hook)) {
81                 DPRINTK("ipt_init_target: check failed for `%s'.\n",
82                         t->u.kernel.target->name);
83                 module_put(t->u.kernel.target->me);
84                 ret = -EINVAL;
85         }
86
87         return ret;
88 }
89
90 static void
91 ipt_destroy_target(struct ipt_entry_target *t)
92 {
93         if (t->u.kernel.target->destroy)
94                 t->u.kernel.target->destroy(t->u.kernel.target, t->data,
95                                             t->u.target_size - sizeof(*t));
96         module_put(t->u.kernel.target->me);
97 }
98
99 static int
100 tcf_ipt_release(struct tcf_ipt *p, int bind)
101 {
102         int ret = 0;
103         if (p) {
104                 if (bind)
105                         p->bindcnt--;
106                 p->refcnt--;
107                 if (p->bindcnt <= 0 && p->refcnt <= 0) {
108                         ipt_destroy_target(p->t);
109                         kfree(p->tname);
110                         kfree(p->t);
111                         tcf_hash_destroy(p);
112                         ret = ACT_P_DELETED;
113                 }
114         }
115         return ret;
116 }
117
118 static int
119 tcf_ipt_init(struct rtattr *rta, struct rtattr *est, struct tc_action *a,
120              int ovr, int bind)
121 {
122         struct rtattr *tb[TCA_IPT_MAX];
123         struct tcf_ipt *p;
124         struct ipt_entry_target *td, *t;
125         char *tname;
126         int ret = 0, err;
127         u32 hook = 0;
128         u32 index = 0;
129
130         if (rta == NULL || rtattr_parse_nested(tb, TCA_IPT_MAX, rta) < 0)
131                 return -EINVAL;
132
133         if (tb[TCA_IPT_HOOK-1] == NULL ||
134             RTA_PAYLOAD(tb[TCA_IPT_HOOK-1]) < sizeof(u32))
135                 return -EINVAL;
136         if (tb[TCA_IPT_TARG-1] == NULL ||
137             RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < sizeof(*t))
138                 return -EINVAL;
139         td = (struct ipt_entry_target *)RTA_DATA(tb[TCA_IPT_TARG-1]);
140         if (RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < td->u.target_size)
141                 return -EINVAL;
142
143         if (tb[TCA_IPT_INDEX-1] != NULL &&
144             RTA_PAYLOAD(tb[TCA_IPT_INDEX-1]) >= sizeof(u32))
145                 index = *(u32 *)RTA_DATA(tb[TCA_IPT_INDEX-1]);
146
147         p = tcf_hash_check(index, a, ovr, bind);
148         if (p == NULL) {
149                 p = tcf_hash_create(index, est, a, sizeof(*p), ovr, bind);
150                 if (p == NULL)
151                         return -ENOMEM;
152                 ret = ACT_P_CREATED;
153         } else {
154                 if (!ovr) {
155                         tcf_ipt_release(p, bind);
156                         return -EEXIST;
157                 }
158         }
159
160         hook = *(u32 *)RTA_DATA(tb[TCA_IPT_HOOK-1]);
161
162         err = -ENOMEM;
163         tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
164         if (tname == NULL)
165                 goto err1;
166         if (tb[TCA_IPT_TABLE - 1] == NULL ||
167             rtattr_strlcpy(tname, tb[TCA_IPT_TABLE-1], IFNAMSIZ) >= IFNAMSIZ)
168                 strcpy(tname, "mangle");
169
170         t = kmalloc(td->u.target_size, GFP_KERNEL);
171         if (t == NULL)
172                 goto err2;
173         memcpy(t, td, td->u.target_size);
174
175         if ((err = ipt_init_target(t, tname, hook)) < 0)
176                 goto err3;
177
178         spin_lock_bh(&p->lock);
179         if (ret != ACT_P_CREATED) {
180                 ipt_destroy_target(p->t);
181                 kfree(p->tname);
182                 kfree(p->t);
183         }
184         p->tname = tname;
185         p->t     = t;
186         p->hook  = hook;
187         spin_unlock_bh(&p->lock);
188         if (ret == ACT_P_CREATED)
189                 tcf_hash_insert(p);
190         return ret;
191
192 err3:
193         kfree(t);
194 err2:
195         kfree(tname);
196 err1:
197         kfree(p);
198         return err;
199 }
200
201 static int
202 tcf_ipt_cleanup(struct tc_action *a, int bind)
203 {
204         struct tcf_ipt *p = PRIV(a, ipt);
205         return tcf_ipt_release(p, bind);
206 }
207
208 static int
209 tcf_ipt(struct sk_buff *skb, struct tc_action *a, struct tcf_result *res)
210 {
211         int ret = 0, result = 0;
212         struct tcf_ipt *p = PRIV(a, ipt);
213
214         if (skb_cloned(skb)) {
215                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
216                         return TC_ACT_UNSPEC;
217         }
218
219         spin_lock(&p->lock);
220
221         p->tm.lastuse = jiffies;
222         p->bstats.bytes += skb->len;
223         p->bstats.packets++;
224
225         /* yes, we have to worry about both in and out dev
226          worry later - danger - this API seems to have changed
227          from earlier kernels */
228
229         /* iptables targets take a double skb pointer in case the skb
230          * needs to be replaced. We don't own the skb, so this must not
231          * happen. The pskb_expand_head above should make sure of this */
232         ret = p->t->u.kernel.target->target(&skb, skb->dev, NULL, p->hook,
233                                             p->t->u.kernel.target, p->t->data,
234                                             NULL);
235         switch (ret) {
236         case NF_ACCEPT:
237                 result = TC_ACT_OK;
238                 break;
239         case NF_DROP:
240                 result = TC_ACT_SHOT;
241                 p->qstats.drops++;
242                 break;
243         case IPT_CONTINUE:
244                 result = TC_ACT_PIPE;
245                 break;
246         default:
247                 if (net_ratelimit())
248                         printk("Bogus netfilter code %d assume ACCEPT\n", ret);
249                 result = TC_POLICE_OK;
250                 break;
251         }
252         spin_unlock(&p->lock);
253         return result;
254
255 }
256
257 static int
258 tcf_ipt_dump(struct sk_buff *skb, struct tc_action *a, int bind, int ref)
259 {
260         struct ipt_entry_target *t;
261         struct tcf_t tm;
262         struct tc_cnt c;
263         unsigned char *b = skb->tail;
264         struct tcf_ipt *p = PRIV(a, ipt);
265
266         /* for simple targets kernel size == user size
267         ** user name = target name
268         ** for foolproof you need to not assume this
269         */
270
271         t = kmalloc(p->t->u.user.target_size, GFP_ATOMIC);
272         if (t == NULL)
273                 goto rtattr_failure;
274
275         c.bindcnt = p->bindcnt - bind;
276         c.refcnt = p->refcnt - ref;
277         memcpy(t, p->t, p->t->u.user.target_size);
278         strcpy(t->u.user.name, p->t->u.kernel.target->name);
279
280         DPRINTK("\ttcf_ipt_dump tablename %s length %d\n", p->tname,
281                 strlen(p->tname));
282         DPRINTK("\tdump target name %s size %d size user %d "
283                 "data[0] %x data[1] %x\n", p->t->u.kernel.target->name,
284                 p->t->u.target_size, p->t->u.user.target_size,
285                 p->t->data[0], p->t->data[1]);
286         RTA_PUT(skb, TCA_IPT_TARG, p->t->u.user.target_size, t);
287         RTA_PUT(skb, TCA_IPT_INDEX, 4, &p->index);
288         RTA_PUT(skb, TCA_IPT_HOOK, 4, &p->hook);
289         RTA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
290         RTA_PUT(skb, TCA_IPT_TABLE, IFNAMSIZ, p->tname);
291         tm.install = jiffies_to_clock_t(jiffies - p->tm.install);
292         tm.lastuse = jiffies_to_clock_t(jiffies - p->tm.lastuse);
293         tm.expires = jiffies_to_clock_t(p->tm.expires);
294         RTA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
295         kfree(t);
296         return skb->len;
297
298       rtattr_failure:
299         skb_trim(skb, b - skb->data);
300         kfree(t);
301         return -1;
302 }
303
304 static struct tc_action_ops act_ipt_ops = {
305         .kind           =       "ipt",
306         .type           =       TCA_ACT_IPT,
307         .capab          =       TCA_CAP_NONE,
308         .owner          =       THIS_MODULE,
309         .act            =       tcf_ipt,
310         .dump           =       tcf_ipt_dump,
311         .cleanup        =       tcf_ipt_cleanup,
312         .lookup         =       tcf_hash_search,
313         .init           =       tcf_ipt_init,
314         .walk           =       tcf_generic_walker
315 };
316
317 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
318 MODULE_DESCRIPTION("Iptables target actions");
319 MODULE_LICENSE("GPL");
320
321 static int __init
322 ipt_init_module(void)
323 {
324         return tcf_register_action(&act_ipt_ops);
325 }
326
327 static void __exit
328 ipt_cleanup_module(void)
329 {
330         tcf_unregister_action(&act_ipt_ops);
331 }
332
333 module_init(ipt_init_module);
334 module_exit(ipt_cleanup_module);