tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
[NETFITLER]: Add nfnetlink layer.
Introduce "nfnetlink" (netfilter netlink) layer. This layer is used as
transport layer for all userspace communication of the new upcoming
netfilter subsystems, such as ctnetlink, nfnetlink_queue and some day even
the mythical pkttables ;)
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
authored by
Harald Welte
and committed by
David S. Miller
f9e815b3
ac3247ba
+497
+145
include/linux/netfilter/nfnetlink.h
+145
include/linux/netfilter/nfnetlink.h
···
1
+
#ifndef _NFNETLINK_H
2
+
#define _NFNETLINK_H
3
+
#include <linux/types.h>
4
+
5
+
/* nfnetlink groups: Up to 32 maximum */
6
+
#define NF_NETLINK_CONNTRACK_NEW 0x00000001
7
+
#define NF_NETLINK_CONNTRACK_UPDATE 0x00000002
8
+
#define NF_NETLINK_CONNTRACK_DESTROY 0x00000004
9
+
#define NF_NETLINK_CONNTRACK_EXP_NEW 0x00000008
10
+
#define NF_NETLINK_CONNTRACK_EXP_UPDATE 0x00000010
11
+
#define NF_NETLINK_CONNTRACK_EXP_DESTROY 0x00000020
12
+
13
+
/* Generic structure for encapsulation optional netfilter information.
14
+
* It is reminiscent of sockaddr, but with sa_family replaced
15
+
* with attribute type.
16
+
* ! This should someday be put somewhere generic as now rtnetlink and
17
+
* ! nfnetlink use the same attributes methods. - J. Schulist.
18
+
*/
19
+
20
+
struct nfattr
21
+
{
22
+
u_int16_t nfa_len;
23
+
u_int16_t nfa_type;
24
+
} __attribute__ ((packed));
25
+
26
+
/* FIXME: Shamelessly copy and pasted from rtnetlink.h, it's time
27
+
* to put this in a generic file */
28
+
29
+
#define NFA_ALIGNTO 4
30
+
#define NFA_ALIGN(len) (((len) + NFA_ALIGNTO - 1) & ~(NFA_ALIGNTO - 1))
31
+
#define NFA_OK(nfa,len) ((len) > 0 && (nfa)->nfa_len >= sizeof(struct nfattr) \
32
+
&& (nfa)->nfa_len <= (len))
33
+
#define NFA_NEXT(nfa,attrlen) ((attrlen) -= NFA_ALIGN((nfa)->nfa_len), \
34
+
(struct nfattr *)(((char *)(nfa)) + NFA_ALIGN((nfa)->nfa_len)))
35
+
#define NFA_LENGTH(len) (NFA_ALIGN(sizeof(struct nfattr)) + (len))
36
+
#define NFA_SPACE(len) NFA_ALIGN(NFA_LENGTH(len))
37
+
#define NFA_DATA(nfa) ((void *)(((char *)(nfa)) + NFA_LENGTH(0)))
38
+
#define NFA_PAYLOAD(nfa) ((int)((nfa)->nfa_len) - NFA_LENGTH(0))
39
+
#define NFA_NEST(skb, type) \
40
+
({ struct nfattr *__start = (struct nfattr *) (skb)->tail; \
41
+
NFA_PUT(skb, type, 0, NULL); \
42
+
__start; })
43
+
#define NFA_NEST_END(skb, start) \
44
+
({ (start)->nfa_len = ((skb)->tail - (unsigned char *) (start)); \
45
+
(skb)->len; })
46
+
#define NFA_NEST_CANCEL(skb, start) \
47
+
({ if (start) \
48
+
skb_trim(skb, (unsigned char *) (start) - (skb)->data); \
49
+
-1; })
50
+
51
+
/* General form of address family dependent message.
52
+
*/
53
+
struct nfgenmsg {
54
+
u_int8_t nfgen_family; /* AF_xxx */
55
+
u_int8_t version; /* nfnetlink version */
56
+
u_int16_t res_id; /* resource id */
57
+
} __attribute__ ((packed));
58
+
59
+
#define NFNETLINK_V1 1
60
+
61
+
#define NFM_NFA(n) ((struct nfattr *)(((char *)(n)) \
62
+
+ NLMSG_ALIGN(sizeof(struct nfgenmsg))))
63
+
#define NFM_PAYLOAD(n) NLMSG_PAYLOAD(n, sizeof(struct nfgenmsg))
64
+
65
+
/* netfilter netlink message types are split in two pieces:
66
+
* 8 bit subsystem, 8bit operation.
67
+
*/
68
+
69
+
#define NFNL_SUBSYS_ID(x) ((x & 0xff00) >> 8)
70
+
#define NFNL_MSG_TYPE(x) (x & 0x00ff)
71
+
72
+
enum nfnl_subsys_id {
73
+
NFNL_SUBSYS_NONE = 0,
74
+
NFNL_SUBSYS_CTNETLINK,
75
+
NFNL_SUBSYS_CTNETLINK_EXP,
76
+
NFNL_SUBSYS_IPTNETLINK,
77
+
NFNL_SUBSYS_QUEUE,
78
+
NFNL_SUBSYS_ULOG,
79
+
NFNL_SUBSYS_COUNT,
80
+
};
81
+
82
+
#ifdef __KERNEL__
83
+
84
+
#include <linux/capability.h>
85
+
86
+
struct nfnl_callback
87
+
{
88
+
kernel_cap_t cap_required; /* capabilities required for this msg */
89
+
int (*call)(struct sock *nl, struct sk_buff *skb,
90
+
struct nlmsghdr *nlh, struct nfattr *cda[], int *errp);
91
+
};
92
+
93
+
struct nfnetlink_subsystem
94
+
{
95
+
const char *name;
96
+
__u8 subsys_id; /* nfnetlink subsystem ID */
97
+
__u8 cb_count; /* number of callbacks */
98
+
u_int32_t attr_count; /* number of nfattr's */
99
+
struct nfnl_callback *cb; /* callback for individual types */
100
+
};
101
+
102
+
extern void __nfa_fill(struct sk_buff *skb, int attrtype,
103
+
int attrlen, const void *data);
104
+
#define NFA_PUT(skb, attrtype, attrlen, data) \
105
+
({ if (skb_tailroom(skb) < (int)NFA_SPACE(attrlen)) goto nfattr_failure; \
106
+
__nfa_fill(skb, attrtype, attrlen, data); })
107
+
108
+
extern struct semaphore nfnl_sem;
109
+
110
+
#define nfnl_shlock() down(&nfnl_sem)
111
+
#define nfnl_shlock_nowait() down_trylock(&nfnl_sem)
112
+
113
+
#define nfnl_shunlock() do { up(&nfnl_sem); \
114
+
if(nfnl && nfnl->sk_receive_queue.qlen) \
115
+
nfnl->sk_data_ready(nfnl, 0); \
116
+
} while(0)
117
+
118
+
extern void nfnl_lock(void);
119
+
extern void nfnl_unlock(void);
120
+
121
+
extern int nfnetlink_subsys_register(struct nfnetlink_subsystem *n);
122
+
extern int nfnetlink_subsys_unregister(struct nfnetlink_subsystem *n);
123
+
124
+
extern int nfattr_parse(struct nfattr *tb[], int maxattr,
125
+
struct nfattr *nfa, int len);
126
+
127
+
#define nfattr_parse_nested(tb, max, nfa) \
128
+
nfattr_parse((tb), (max), NFA_DATA((nfa)), NFA_PAYLOAD((nfa)))
129
+
130
+
#define nfattr_bad_size(tb, max, cta_min) \
131
+
({ int __i, __res = 0; \
132
+
for (__i=0; __i<max; __i++) \
133
+
if (tb[__i] && NFA_PAYLOAD(tb[__i]) < cta_min[__i]){ \
134
+
__res = 1; \
135
+
break; \
136
+
} \
137
+
__res; \
138
+
})
139
+
140
+
extern int nfnetlink_send(struct sk_buff *skb, u32 pid, unsigned group,
141
+
int echo);
142
+
extern int nfnetlink_unicast(struct sk_buff *skb, u_int32_t pid, int flags);
143
+
144
+
#endif /* __KERNEL__ */
145
+
#endif /* _NFNETLINK_H */
+2
net/Kconfig
+2
net/Kconfig
+1
net/Makefile
+1
net/Makefile
+5
net/netfilter/Kconfig
+5
net/netfilter/Kconfig
+1
net/netfilter/Makefile
+1
net/netfilter/Makefile
···
1
+
obj-$(CONFIG_NETFILTER_NETLINK) += nfnetlink.o
+343
net/netfilter/nfnetlink.c
+343
net/netfilter/nfnetlink.c
···
1
+
/* Netfilter messages via netlink socket. Allows for user space
2
+
* protocol helpers and general trouble making from userspace.
3
+
*
4
+
* (C) 2001 by Jay Schulist <jschlst@samba.org>,
5
+
* (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
6
+
* (C) 2005 by Pablo Neira Ayuso <pablo@eurodev.net>
7
+
*
8
+
* Initial netfilter messages via netlink development funded and
9
+
* generally made possible by Network Robots, Inc. (www.networkrobots.com)
10
+
*
11
+
* Further development of this code funded by Astaro AG (http://www.astaro.com)
12
+
*
13
+
* This software may be used and distributed according to the terms
14
+
* of the GNU General Public License, incorporated herein by reference.
15
+
*/
16
+
17
+
#include <linux/config.h>
18
+
#include <linux/module.h>
19
+
#include <linux/types.h>
20
+
#include <linux/socket.h>
21
+
#include <linux/kernel.h>
22
+
#include <linux/major.h>
23
+
#include <linux/sched.h>
24
+
#include <linux/timer.h>
25
+
#include <linux/string.h>
26
+
#include <linux/sockios.h>
27
+
#include <linux/net.h>
28
+
#include <linux/fcntl.h>
29
+
#include <linux/skbuff.h>
30
+
#include <asm/uaccess.h>
31
+
#include <asm/system.h>
32
+
#include <net/sock.h>
33
+
#include <linux/init.h>
34
+
#include <linux/spinlock.h>
35
+
36
+
#include <linux/netfilter.h>
37
+
#include <linux/netlink.h>
38
+
#include <linux/netfilter/nfnetlink.h>
39
+
40
+
MODULE_LICENSE("GPL");
41
+
42
+
static char __initdata nfversion[] = "0.30";
43
+
44
+
#if 0
45
+
#define DEBUGP printk
46
+
#else
47
+
#define DEBUGP(format, args...)
48
+
#endif
49
+
50
+
static struct sock *nfnl = NULL;
51
+
static struct nfnetlink_subsystem *subsys_table[NFNL_SUBSYS_COUNT];
52
+
DECLARE_MUTEX(nfnl_sem);
53
+
54
+
void nfnl_lock(void)
55
+
{
56
+
nfnl_shlock();
57
+
}
58
+
59
+
void nfnl_unlock(void)
60
+
{
61
+
nfnl_shunlock();
62
+
}
63
+
64
+
int nfnetlink_subsys_register(struct nfnetlink_subsystem *n)
65
+
{
66
+
DEBUGP("registering subsystem ID %u\n", n->subsys_id);
67
+
68
+
/* If the netlink socket wasn't created, then fail */
69
+
if (!nfnl)
70
+
return -1;
71
+
72
+
nfnl_lock();
73
+
subsys_table[n->subsys_id] = n;
74
+
nfnl_unlock();
75
+
76
+
return 0;
77
+
}
78
+
79
+
int nfnetlink_subsys_unregister(struct nfnetlink_subsystem *n)
80
+
{
81
+
DEBUGP("unregistering subsystem ID %u\n", n->subsys_id);
82
+
83
+
nfnl_lock();
84
+
subsys_table[n->subsys_id] = NULL;
85
+
nfnl_unlock();
86
+
87
+
return 0;
88
+
}
89
+
90
+
static inline struct nfnetlink_subsystem *nfnetlink_get_subsys(u_int16_t type)
91
+
{
92
+
u_int8_t subsys_id = NFNL_SUBSYS_ID(type);
93
+
94
+
if (subsys_id >= NFNL_SUBSYS_COUNT
95
+
|| subsys_table[subsys_id] == NULL)
96
+
return NULL;
97
+
98
+
return subsys_table[subsys_id];
99
+
}
100
+
101
+
static inline struct nfnl_callback *
102
+
nfnetlink_find_client(u_int16_t type, struct nfnetlink_subsystem *ss)
103
+
{
104
+
u_int8_t cb_id = NFNL_MSG_TYPE(type);
105
+
106
+
if (cb_id >= ss->cb_count) {
107
+
DEBUGP("msgtype %u >= %u, returning\n", type, ss->cb_count);
108
+
return NULL;
109
+
}
110
+
111
+
return &ss->cb[cb_id];
112
+
}
113
+
114
+
void __nfa_fill(struct sk_buff *skb, int attrtype, int attrlen,
115
+
const void *data)
116
+
{
117
+
struct nfattr *nfa;
118
+
int size = NFA_LENGTH(attrlen);
119
+
120
+
nfa = (struct nfattr *)skb_put(skb, NFA_ALIGN(size));
121
+
nfa->nfa_type = attrtype;
122
+
nfa->nfa_len = size;
123
+
memcpy(NFA_DATA(nfa), data, attrlen);
124
+
}
125
+
126
+
int nfattr_parse(struct nfattr *tb[], int maxattr, struct nfattr *nfa, int len)
127
+
{
128
+
memset(tb, 0, sizeof(struct nfattr *) * maxattr);
129
+
130
+
while (NFA_OK(nfa, len)) {
131
+
unsigned flavor = nfa->nfa_type;
132
+
if (flavor && flavor <= maxattr)
133
+
tb[flavor-1] = nfa;
134
+
nfa = NFA_NEXT(nfa, len);
135
+
}
136
+
137
+
return 0;
138
+
}
139
+
140
+
/**
141
+
* nfnetlink_check_attributes - check and parse nfnetlink attributes
142
+
*
143
+
* subsys: nfnl subsystem for which this message is to be parsed
144
+
* nlmsghdr: netlink message to be checked/parsed
145
+
* cda: array of pointers, needs to be at least subsys->attr_count big
146
+
*
147
+
*/
148
+
static int
149
+
nfnetlink_check_attributes(struct nfnetlink_subsystem *subsys,
150
+
struct nlmsghdr *nlh, struct nfattr *cda[])
151
+
{
152
+
int min_len;
153
+
154
+
memset(cda, 0, sizeof(struct nfattr *) * subsys->attr_count);
155
+
156
+
/* check attribute lengths. */
157
+
min_len = NLMSG_ALIGN(sizeof(struct nfgenmsg));
158
+
if (nlh->nlmsg_len < min_len)
159
+
return -EINVAL;
160
+
161
+
if (nlh->nlmsg_len > min_len) {
162
+
struct nfattr *attr = NFM_NFA(NLMSG_DATA(nlh));
163
+
int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
164
+
165
+
while (NFA_OK(attr, attrlen)) {
166
+
unsigned flavor = attr->nfa_type;
167
+
if (flavor) {
168
+
if (flavor > subsys->attr_count)
169
+
return -EINVAL;
170
+
cda[flavor - 1] = attr;
171
+
}
172
+
attr = NFA_NEXT(attr, attrlen);
173
+
}
174
+
} else
175
+
return -EINVAL;
176
+
177
+
return 0;
178
+
}
179
+
180
+
int nfnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
181
+
{
182
+
int allocation = in_interrupt() ? GFP_ATOMIC : GFP_KERNEL;
183
+
int err = 0;
184
+
185
+
NETLINK_CB(skb).dst_groups = group;
186
+
if (echo)
187
+
atomic_inc(&skb->users);
188
+
netlink_broadcast(nfnl, skb, pid, group, allocation);
189
+
if (echo)
190
+
err = netlink_unicast(nfnl, skb, pid, MSG_DONTWAIT);
191
+
192
+
return err;
193
+
}
194
+
195
+
int nfnetlink_unicast(struct sk_buff *skb, u_int32_t pid, int flags)
196
+
{
197
+
return netlink_unicast(nfnl, skb, pid, flags);
198
+
}
199
+
200
+
/* Process one complete nfnetlink message. */
201
+
static inline int nfnetlink_rcv_msg(struct sk_buff *skb,
202
+
struct nlmsghdr *nlh, int *errp)
203
+
{
204
+
struct nfnl_callback *nc;
205
+
struct nfnetlink_subsystem *ss;
206
+
int type, err = 0;
207
+
208
+
DEBUGP("entered; subsys=%u, msgtype=%u\n",
209
+
NFNL_SUBSYS_ID(nlh->nlmsg_type),
210
+
NFNL_MSG_TYPE(nlh->nlmsg_type));
211
+
212
+
/* Only requests are handled by kernel now. */
213
+
if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) {
214
+
DEBUGP("received non-request message\n");
215
+
return 0;
216
+
}
217
+
218
+
/* All the messages must at least contain nfgenmsg */
219
+
if (nlh->nlmsg_len <
220
+
NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg)))) {
221
+
DEBUGP("received message was too short\n");
222
+
return 0;
223
+
}
224
+
225
+
type = nlh->nlmsg_type;
226
+
ss = nfnetlink_get_subsys(type);
227
+
if (!ss)
228
+
goto err_inval;
229
+
230
+
nc = nfnetlink_find_client(type, ss);
231
+
if (!nc) {
232
+
DEBUGP("unable to find client for type %d\n", type);
233
+
goto err_inval;
234
+
}
235
+
236
+
if (nc->cap_required &&
237
+
!cap_raised(NETLINK_CB(skb).eff_cap, nc->cap_required)) {
238
+
DEBUGP("permission denied for type %d\n", type);
239
+
*errp = -EPERM;
240
+
return -1;
241
+
}
242
+
243
+
{
244
+
struct nfattr *cda[ss->attr_count];
245
+
246
+
memset(cda, 0, ss->attr_count*sizeof(struct nfattr *));
247
+
248
+
err = nfnetlink_check_attributes(ss, nlh, cda);
249
+
if (err < 0)
250
+
goto err_inval;
251
+
252
+
err = nc->call(nfnl, skb, nlh, cda, errp);
253
+
*errp = err;
254
+
return err;
255
+
}
256
+
257
+
err_inval:
258
+
*errp = -EINVAL;
259
+
return -1;
260
+
}
261
+
262
+
/* Process one packet of messages. */
263
+
static inline int nfnetlink_rcv_skb(struct sk_buff *skb)
264
+
{
265
+
int err;
266
+
struct nlmsghdr *nlh;
267
+
268
+
while (skb->len >= NLMSG_SPACE(0)) {
269
+
u32 rlen;
270
+
271
+
nlh = (struct nlmsghdr *)skb->data;
272
+
if (nlh->nlmsg_len < sizeof(struct nlmsghdr)
273
+
|| skb->len < nlh->nlmsg_len)
274
+
return 0;
275
+
rlen = NLMSG_ALIGN(nlh->nlmsg_len);
276
+
if (rlen > skb->len)
277
+
rlen = skb->len;
278
+
if (nfnetlink_rcv_msg(skb, nlh, &err)) {
279
+
if (!err)
280
+
return -1;
281
+
netlink_ack(skb, nlh, err);
282
+
} else
283
+
if (nlh->nlmsg_flags & NLM_F_ACK)
284
+
netlink_ack(skb, nlh, 0);
285
+
skb_pull(skb, rlen);
286
+
}
287
+
288
+
return 0;
289
+
}
290
+
291
+
static void nfnetlink_rcv(struct sock *sk, int len)
292
+
{
293
+
do {
294
+
struct sk_buff *skb;
295
+
296
+
if (nfnl_shlock_nowait())
297
+
return;
298
+
299
+
while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
300
+
if (nfnetlink_rcv_skb(skb)) {
301
+
if (skb->len)
302
+
skb_queue_head(&sk->sk_receive_queue,
303
+
skb);
304
+
else
305
+
kfree_skb(skb);
306
+
break;
307
+
}
308
+
kfree_skb(skb);
309
+
}
310
+
311
+
up(&nfnl_sem);
312
+
} while(nfnl && nfnl->sk_receive_queue.qlen);
313
+
}
314
+
315
+
void __exit nfnetlink_exit(void)
316
+
{
317
+
printk("Removing netfilter NETLINK layer.\n");
318
+
sock_release(nfnl->sk_socket);
319
+
return;
320
+
}
321
+
322
+
int __init nfnetlink_init(void)
323
+
{
324
+
printk("Netfilter messages via NETLINK v%s.\n", nfversion);
325
+
326
+
nfnl = netlink_kernel_create(NETLINK_NETFILTER, nfnetlink_rcv);
327
+
if (!nfnl) {
328
+
printk(KERN_ERR "cannot initialize nfnetlink!\n");
329
+
return -1;
330
+
}
331
+
332
+
return 0;
333
+
}
334
+
335
+
module_init(nfnetlink_init);
336
+
module_exit(nfnetlink_exit);
337
+
338
+
EXPORT_SYMBOL_GPL(nfnetlink_subsys_register);
339
+
EXPORT_SYMBOL_GPL(nfnetlink_subsys_unregister);
340
+
EXPORT_SYMBOL_GPL(nfnetlink_send);
341
+
EXPORT_SYMBOL_GPL(nfnetlink_unicast);
342
+
EXPORT_SYMBOL_GPL(nfattr_parse);
343
+
EXPORT_SYMBOL_GPL(__nfa_fill);