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