Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[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 <asm/uaccess.h>
15 #include <asm/system.h>
16 #include <asm/bitops.h>
17 #include <linux/config.h>
18 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/sched.h>
21 #include <linux/string.h>
22 #include <linux/mm.h>
23 #include <linux/socket.h>
24 #include <linux/sockios.h>
25 #include <linux/in.h>
26 #include <linux/errno.h>
27 #include <linux/interrupt.h>
28 #include <linux/netdevice.h>
29 #include <linux/skbuff.h>
30 #include <linux/rtnetlink.h>
31 #include <linux/module.h>
32 #include <linux/init.h>
33 #include <linux/proc_fs.h>
34 #include <linux/kmod.h>
35 #include <net/sock.h>
36 #include <net/pkt_sched.h>
37 #include <linux/tc_act/tc_ipt.h>
38 #include <net/tc_act/tc_ipt.h>
39
40 #include <linux/netfilter_ipv4/ip_tables.h>
41
42 /* use generic hash table */
43 #define MY_TAB_SIZE     16
44 #define MY_TAB_MASK     15
45
46 static u32 idx_gen;
47 static struct tcf_ipt *tcf_ipt_ht[MY_TAB_SIZE];
48 /* ipt hash table lock */
49 static DEFINE_RWLOCK(ipt_lock);
50
51 /* ovewrride the defaults */
52 #define tcf_st          tcf_ipt
53 #define tcf_t_lock      ipt_lock
54 #define tcf_ht          tcf_ipt_ht
55
56 #define CONFIG_NET_ACT_INIT
57 #include <net/pkt_act.h>
58
59 static int
60 ipt_init_target(struct ipt_entry_target *t, char *table, unsigned int hook)
61 {
62         struct ipt_target *target;
63         int ret = 0;
64
65         target = xt_find_target(AF_INET, t->u.user.name, t->u.user.revision);
66         if (!target)
67                 return -ENOENT;
68
69         DPRINTK("ipt_init_target: found %s\n", target->name);
70         t->u.kernel.target = target;
71
72         if (t->u.kernel.target->checkentry
73             && !t->u.kernel.target->checkentry(table, NULL,
74                                                t->u.kernel.target, t->data,
75                                                t->u.target_size - sizeof(*t),
76                                                hook)) {
77                 DPRINTK("ipt_init_target: check failed for `%s'.\n",
78                         t->u.kernel.target->name);
79                 module_put(t->u.kernel.target->me);
80                 ret = -EINVAL;
81         }
82
83         return ret;
84 }
85
86 static void
87 ipt_destroy_target(struct ipt_entry_target *t)
88 {
89         if (t->u.kernel.target->destroy)
90                 t->u.kernel.target->destroy(t->u.kernel.target, t->data,
91                                             t->u.target_size - sizeof(*t));
92         module_put(t->u.kernel.target->me);
93 }
94
95 static int
96 tcf_ipt_release(struct tcf_ipt *p, int bind)
97 {
98         int ret = 0;
99         if (p) {
100                 if (bind)
101                         p->bindcnt--;
102                 p->refcnt--;
103                 if (p->bindcnt <= 0 && p->refcnt <= 0) {
104                         ipt_destroy_target(p->t);
105                         kfree(p->tname);
106                         kfree(p->t);
107                         tcf_hash_destroy(p);
108                         ret = ACT_P_DELETED;
109                 }
110         }
111         return ret;
112 }
113
114 static int
115 tcf_ipt_init(struct rtattr *rta, struct rtattr *est, struct tc_action *a,
116              int ovr, int bind)
117 {
118         struct rtattr *tb[TCA_IPT_MAX];
119         struct tcf_ipt *p;
120         struct ipt_entry_target *td, *t;
121         char *tname;
122         int ret = 0, err;
123         u32 hook = 0;
124         u32 index = 0;
125
126         if (rta == NULL || rtattr_parse_nested(tb, TCA_IPT_MAX, rta) < 0)
127                 return -EINVAL;
128
129         if (tb[TCA_IPT_HOOK-1] == NULL ||
130             RTA_PAYLOAD(tb[TCA_IPT_HOOK-1]) < sizeof(u32))
131                 return -EINVAL;
132         if (tb[TCA_IPT_TARG-1] == NULL ||
133             RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < sizeof(*t))
134                 return -EINVAL;
135         td = (struct ipt_entry_target *)RTA_DATA(tb[TCA_IPT_TARG-1]);
136         if (RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < td->u.target_size)
137                 return -EINVAL;
138
139         if (tb[TCA_IPT_INDEX-1] != NULL &&
140             RTA_PAYLOAD(tb[TCA_IPT_INDEX-1]) >= sizeof(u32))
141                 index = *(u32 *)RTA_DATA(tb[TCA_IPT_INDEX-1]);
142
143         p = tcf_hash_check(index, a, ovr, bind);
144         if (p == NULL) {
145                 p = tcf_hash_create(index, est, a, sizeof(*p), ovr, bind);
146                 if (p == NULL)
147                         return -ENOMEM;
148                 ret = ACT_P_CREATED;
149         } else {
150                 if (!ovr) {
151                         tcf_ipt_release(p, bind);
152                         return -EEXIST;
153                 }
154         }
155
156         hook = *(u32 *)RTA_DATA(tb[TCA_IPT_HOOK-1]);
157
158         err = -ENOMEM;
159         tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
160         if (tname == NULL)
161                 goto err1;
162         if (tb[TCA_IPT_TABLE - 1] == NULL ||
163             rtattr_strlcpy(tname, tb[TCA_IPT_TABLE-1], IFNAMSIZ) >= IFNAMSIZ)
164                 strcpy(tname, "mangle");
165
166         t = kmalloc(td->u.target_size, GFP_KERNEL);
167         if (t == NULL)
168                 goto err2;
169         memcpy(t, td, td->u.target_size);
170
171         if ((err = ipt_init_target(t, tname, hook)) < 0)
172                 goto err3;
173
174         spin_lock_bh(&p->lock);
175         if (ret != ACT_P_CREATED) {
176                 ipt_destroy_target(p->t);
177                 kfree(p->tname);
178                 kfree(p->t);
179         }
180         p->tname = tname;
181         p->t     = t;
182         p->hook  = hook;
183         spin_unlock_bh(&p->lock);
184         if (ret == ACT_P_CREATED)
185                 tcf_hash_insert(p);
186         return ret;
187
188 err3:
189         kfree(t);
190 err2:
191         kfree(tname);
192 err1:
193         kfree(p);
194         return err;
195 }
196
197 static int
198 tcf_ipt_cleanup(struct tc_action *a, int bind)
199 {
200         struct tcf_ipt *p = PRIV(a, ipt);
201         return tcf_ipt_release(p, bind);
202 }
203
204 static int
205 tcf_ipt(struct sk_buff *skb, struct tc_action *a, struct tcf_result *res)
206 {
207         int ret = 0, result = 0;
208         struct tcf_ipt *p = PRIV(a, ipt);
209
210         if (skb_cloned(skb)) {
211                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
212                         return TC_ACT_UNSPEC;
213         }
214
215         spin_lock(&p->lock);
216
217         p->tm.lastuse = jiffies;
218         p->bstats.bytes += skb->len;
219         p->bstats.packets++;
220
221         /* yes, we have to worry about both in and out dev
222          worry later - danger - this API seems to have changed
223          from earlier kernels */
224
225         /* iptables targets take a double skb pointer in case the skb
226          * needs to be replaced. We don't own the skb, so this must not
227          * happen. The pskb_expand_head above should make sure of this */
228         ret = p->t->u.kernel.target->target(&skb, skb->dev, NULL, p->hook,
229                                             p->t->u.kernel.target, p->t->data,
230                                             NULL);
231         switch (ret) {
232         case NF_ACCEPT:
233                 result = TC_ACT_OK;
234                 break;
235         case NF_DROP:
236                 result = TC_ACT_SHOT;
237                 p->qstats.drops++;
238                 break;
239         case IPT_CONTINUE:
240                 result = TC_ACT_PIPE;
241                 break;
242         default:
243                 if (net_ratelimit())
244                         printk("Bogus netfilter code %d assume ACCEPT\n", ret);
245                 result = TC_POLICE_OK;
246                 break;
247         }
248         spin_unlock(&p->lock);
249         return result;
250
251 }
252
253 static int
254 tcf_ipt_dump(struct sk_buff *skb, struct tc_action *a, int bind, int ref)
255 {
256         struct ipt_entry_target *t;
257         struct tcf_t tm;
258         struct tc_cnt c;
259         unsigned char *b = skb->tail;
260         struct tcf_ipt *p = PRIV(a, ipt);
261
262         /* for simple targets kernel size == user size
263         ** user name = target name
264         ** for foolproof you need to not assume this
265         */
266
267         t = kmalloc(p->t->u.user.target_size, GFP_ATOMIC);
268         if (t == NULL)
269                 goto rtattr_failure;
270
271         c.bindcnt = p->bindcnt - bind;
272         c.refcnt = p->refcnt - ref;
273         memcpy(t, p->t, p->t->u.user.target_size);
274         strcpy(t->u.user.name, p->t->u.kernel.target->name);
275
276         DPRINTK("\ttcf_ipt_dump tablename %s length %d\n", p->tname,
277                 strlen(p->tname));
278         DPRINTK("\tdump target name %s size %d size user %d "
279                 "data[0] %x data[1] %x\n", p->t->u.kernel.target->name,
280                 p->t->u.target_size, p->t->u.user.target_size,
281                 p->t->data[0], p->t->data[1]);
282         RTA_PUT(skb, TCA_IPT_TARG, p->t->u.user.target_size, t);
283         RTA_PUT(skb, TCA_IPT_INDEX, 4, &p->index);
284         RTA_PUT(skb, TCA_IPT_HOOK, 4, &p->hook);
285         RTA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
286         RTA_PUT(skb, TCA_IPT_TABLE, IFNAMSIZ, p->tname);
287         tm.install = jiffies_to_clock_t(jiffies - p->tm.install);
288         tm.lastuse = jiffies_to_clock_t(jiffies - p->tm.lastuse);
289         tm.expires = jiffies_to_clock_t(p->tm.expires);
290         RTA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
291         kfree(t);
292         return skb->len;
293
294       rtattr_failure:
295         skb_trim(skb, b - skb->data);
296         kfree(t);
297         return -1;
298 }
299
300 static struct tc_action_ops act_ipt_ops = {
301         .kind           =       "ipt",
302         .type           =       TCA_ACT_IPT,
303         .capab          =       TCA_CAP_NONE,
304         .owner          =       THIS_MODULE,
305         .act            =       tcf_ipt,
306         .dump           =       tcf_ipt_dump,
307         .cleanup        =       tcf_ipt_cleanup,
308         .lookup         =       tcf_hash_search,
309         .init           =       tcf_ipt_init,
310         .walk           =       tcf_generic_walker
311 };
312
313 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
314 MODULE_DESCRIPTION("Iptables target actions");
315 MODULE_LICENSE("GPL");
316
317 static int __init
318 ipt_init_module(void)
319 {
320         return tcf_register_action(&act_ipt_ops);
321 }
322
323 static void __exit
324 ipt_cleanup_module(void)
325 {
326         tcf_unregister_action(&act_ipt_ops);
327 }
328
329 module_init(ipt_init_module);
330 module_exit(ipt_cleanup_module);