net_sched: factorize qdisc stats handling
[linux-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 <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 <linux/slab.h>
23 #include <net/netlink.h>
24 #include <net/pkt_sched.h>
25 #include <linux/tc_act/tc_ipt.h>
26 #include <net/tc_act/tc_ipt.h>
27
28 #include <linux/netfilter_ipv4/ip_tables.h>
29
30
31 #define IPT_TAB_MASK     15
32 static struct tcf_common *tcf_ipt_ht[IPT_TAB_MASK + 1];
33 static u32 ipt_idx_gen;
34 static DEFINE_RWLOCK(ipt_lock);
35
36 static struct tcf_hashinfo ipt_hash_info = {
37         .htab   =       tcf_ipt_ht,
38         .hmask  =       IPT_TAB_MASK,
39         .lock   =       &ipt_lock,
40 };
41
42 static int ipt_init_target(struct xt_entry_target *t, char *table, unsigned int hook)
43 {
44         struct xt_tgchk_param par;
45         struct xt_target *target;
46         int ret = 0;
47
48         target = xt_request_find_target(AF_INET, t->u.user.name,
49                                         t->u.user.revision);
50         if (IS_ERR(target))
51                 return PTR_ERR(target);
52
53         t->u.kernel.target = target;
54         par.table     = table;
55         par.entryinfo = NULL;
56         par.target    = target;
57         par.targinfo  = t->data;
58         par.hook_mask = hook;
59         par.family    = NFPROTO_IPV4;
60
61         ret = xt_check_target(&par, t->u.target_size - sizeof(*t), 0, false);
62         if (ret < 0) {
63                 module_put(t->u.kernel.target->me);
64                 return ret;
65         }
66         return 0;
67 }
68
69 static void ipt_destroy_target(struct xt_entry_target *t)
70 {
71         struct xt_tgdtor_param par = {
72                 .target   = t->u.kernel.target,
73                 .targinfo = t->data,
74         };
75         if (par.target->destroy != NULL)
76                 par.target->destroy(&par);
77         module_put(par.target->me);
78 }
79
80 static int tcf_ipt_release(struct tcf_ipt *ipt, int bind)
81 {
82         int ret = 0;
83         if (ipt) {
84                 if (bind)
85                         ipt->tcf_bindcnt--;
86                 ipt->tcf_refcnt--;
87                 if (ipt->tcf_bindcnt <= 0 && ipt->tcf_refcnt <= 0) {
88                         ipt_destroy_target(ipt->tcfi_t);
89                         kfree(ipt->tcfi_tname);
90                         kfree(ipt->tcfi_t);
91                         tcf_hash_destroy(&ipt->common, &ipt_hash_info);
92                         ret = ACT_P_DELETED;
93                 }
94         }
95         return ret;
96 }
97
98 static const struct nla_policy ipt_policy[TCA_IPT_MAX + 1] = {
99         [TCA_IPT_TABLE] = { .type = NLA_STRING, .len = IFNAMSIZ },
100         [TCA_IPT_HOOK]  = { .type = NLA_U32 },
101         [TCA_IPT_INDEX] = { .type = NLA_U32 },
102         [TCA_IPT_TARG]  = { .len = sizeof(struct xt_entry_target) },
103 };
104
105 static int tcf_ipt_init(struct nlattr *nla, struct nlattr *est,
106                         struct tc_action *a, int ovr, int bind)
107 {
108         struct nlattr *tb[TCA_IPT_MAX + 1];
109         struct tcf_ipt *ipt;
110         struct tcf_common *pc;
111         struct xt_entry_target *td, *t;
112         char *tname;
113         int ret = 0, err;
114         u32 hook = 0;
115         u32 index = 0;
116
117         if (nla == NULL)
118                 return -EINVAL;
119
120         err = nla_parse_nested(tb, TCA_IPT_MAX, nla, ipt_policy);
121         if (err < 0)
122                 return err;
123
124         if (tb[TCA_IPT_HOOK] == NULL)
125                 return -EINVAL;
126         if (tb[TCA_IPT_TARG] == NULL)
127                 return -EINVAL;
128
129         td = (struct xt_entry_target *)nla_data(tb[TCA_IPT_TARG]);
130         if (nla_len(tb[TCA_IPT_TARG]) < td->u.target_size)
131                 return -EINVAL;
132
133         if (tb[TCA_IPT_INDEX] != NULL)
134                 index = nla_get_u32(tb[TCA_IPT_INDEX]);
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 (IS_ERR(pc))
141                     return PTR_ERR(pc);
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 = nla_get_u32(tb[TCA_IPT_HOOK]);
152
153         err = -ENOMEM;
154         tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
155         if (unlikely(!tname))
156                 goto err1;
157         if (tb[TCA_IPT_TABLE] == NULL ||
158             nla_strlcpy(tname, tb[TCA_IPT_TABLE], 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         struct xt_action_param par;
203
204         if (skb_cloned(skb)) {
205                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
206                         return TC_ACT_UNSPEC;
207         }
208
209         spin_lock(&ipt->tcf_lock);
210
211         ipt->tcf_tm.lastuse = jiffies;
212         bstats_update(&ipt->tcf_bstats, skb);
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 XT_CONTINUE:
233                 result = TC_ACT_PIPE;
234                 break;
235         default:
236                 if (net_ratelimit())
237                         pr_notice("tc filter: Bogus netfilter code"
238                                   " %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_pointer(skb);
250         struct tcf_ipt *ipt = a->priv;
251         struct xt_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 nla_put_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         NLA_PUT(skb, TCA_IPT_TARG, ipt->tcfi_t->u.user.target_size, t);
269         NLA_PUT_U32(skb, TCA_IPT_INDEX, ipt->tcf_index);
270         NLA_PUT_U32(skb, TCA_IPT_HOOK, ipt->tcfi_hook);
271         NLA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
272         NLA_PUT_STRING(skb, TCA_IPT_TABLE, 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         NLA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
277         kfree(t);
278         return skb->len;
279
280 nla_put_failure:
281         nlmsg_trim(skb, b);
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);