2 * Linux Socket Filter - Kernel level socket filtering
5 * Jay Schulist <jschlst@samba.org>
7 * Based on the design of:
8 * - The Berkeley Packet Filter
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
15 * Andi Kleen - Fix a few bad bugs and races.
16 * Kris Katterjohn - Added many additional checks in sk_chk_filter()
19 #include <linux/module.h>
20 #include <linux/types.h>
22 #include <linux/fcntl.h>
23 #include <linux/socket.h>
25 #include <linux/inet.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_packet.h>
29 #include <net/protocol.h>
30 #include <linux/skbuff.h>
32 #include <linux/errno.h>
33 #include <linux/timer.h>
34 #include <asm/system.h>
35 #include <asm/uaccess.h>
36 #include <asm/unaligned.h>
37 #include <linux/filter.h>
39 /* No hurry in this branch */
40 static void *__load_pointer(struct sk_buff
*skb
, int k
)
45 ptr
= skb_network_header(skb
) + k
- SKF_NET_OFF
;
46 else if (k
>= SKF_LL_OFF
)
47 ptr
= skb_mac_header(skb
) + k
- SKF_LL_OFF
;
49 if (ptr
>= skb
->head
&& ptr
< skb_tail_pointer(skb
))
54 static inline void *load_pointer(struct sk_buff
*skb
, int k
,
55 unsigned int size
, void *buffer
)
58 return skb_header_pointer(skb
, k
, size
, buffer
);
62 return __load_pointer(skb
, k
);
67 * sk_run_filter - run a filter on a socket
68 * @skb: buffer to run the filter on
69 * @filter: filter to apply
70 * @flen: length of filter
72 * Decode and apply filter instructions to the skb->data.
73 * Return length to keep, 0 for none. skb is the data we are
74 * filtering, filter is the array of filter instructions, and
75 * len is the number of filter blocks in the array.
77 unsigned int sk_run_filter(struct sk_buff
*skb
, struct sock_filter
*filter
, int flen
)
79 struct sock_filter
*fentry
; /* We walk down these */
81 u32 A
= 0; /* Accumulator */
82 u32 X
= 0; /* Index Register */
83 u32 mem
[BPF_MEMWORDS
]; /* Scratch Memory Store */
89 * Process array of filter instructions.
91 for (pc
= 0; pc
< flen
; pc
++) {
94 switch (fentry
->code
) {
95 case BPF_ALU
|BPF_ADD
|BPF_X
:
98 case BPF_ALU
|BPF_ADD
|BPF_K
:
101 case BPF_ALU
|BPF_SUB
|BPF_X
:
104 case BPF_ALU
|BPF_SUB
|BPF_K
:
107 case BPF_ALU
|BPF_MUL
|BPF_X
:
110 case BPF_ALU
|BPF_MUL
|BPF_K
:
113 case BPF_ALU
|BPF_DIV
|BPF_X
:
118 case BPF_ALU
|BPF_DIV
|BPF_K
:
121 case BPF_ALU
|BPF_AND
|BPF_X
:
124 case BPF_ALU
|BPF_AND
|BPF_K
:
127 case BPF_ALU
|BPF_OR
|BPF_X
:
130 case BPF_ALU
|BPF_OR
|BPF_K
:
133 case BPF_ALU
|BPF_LSH
|BPF_X
:
136 case BPF_ALU
|BPF_LSH
|BPF_K
:
139 case BPF_ALU
|BPF_RSH
|BPF_X
:
142 case BPF_ALU
|BPF_RSH
|BPF_K
:
145 case BPF_ALU
|BPF_NEG
:
151 case BPF_JMP
|BPF_JGT
|BPF_K
:
152 pc
+= (A
> fentry
->k
) ? fentry
->jt
: fentry
->jf
;
154 case BPF_JMP
|BPF_JGE
|BPF_K
:
155 pc
+= (A
>= fentry
->k
) ? fentry
->jt
: fentry
->jf
;
157 case BPF_JMP
|BPF_JEQ
|BPF_K
:
158 pc
+= (A
== fentry
->k
) ? fentry
->jt
: fentry
->jf
;
160 case BPF_JMP
|BPF_JSET
|BPF_K
:
161 pc
+= (A
& fentry
->k
) ? fentry
->jt
: fentry
->jf
;
163 case BPF_JMP
|BPF_JGT
|BPF_X
:
164 pc
+= (A
> X
) ? fentry
->jt
: fentry
->jf
;
166 case BPF_JMP
|BPF_JGE
|BPF_X
:
167 pc
+= (A
>= X
) ? fentry
->jt
: fentry
->jf
;
169 case BPF_JMP
|BPF_JEQ
|BPF_X
:
170 pc
+= (A
== X
) ? fentry
->jt
: fentry
->jf
;
172 case BPF_JMP
|BPF_JSET
|BPF_X
:
173 pc
+= (A
& X
) ? fentry
->jt
: fentry
->jf
;
175 case BPF_LD
|BPF_W
|BPF_ABS
:
178 ptr
= load_pointer(skb
, k
, 4, &tmp
);
180 A
= ntohl(get_unaligned((__be32
*)ptr
));
184 case BPF_LD
|BPF_H
|BPF_ABS
:
187 ptr
= load_pointer(skb
, k
, 2, &tmp
);
189 A
= ntohs(get_unaligned((__be16
*)ptr
));
193 case BPF_LD
|BPF_B
|BPF_ABS
:
196 ptr
= load_pointer(skb
, k
, 1, &tmp
);
202 case BPF_LD
|BPF_W
|BPF_LEN
:
205 case BPF_LDX
|BPF_W
|BPF_LEN
:
208 case BPF_LD
|BPF_W
|BPF_IND
:
211 case BPF_LD
|BPF_H
|BPF_IND
:
214 case BPF_LD
|BPF_B
|BPF_IND
:
217 case BPF_LDX
|BPF_B
|BPF_MSH
:
218 ptr
= load_pointer(skb
, fentry
->k
, 1, &tmp
);
220 X
= (*(u8
*)ptr
& 0xf) << 2;
227 case BPF_LDX
|BPF_IMM
:
233 case BPF_LDX
|BPF_MEM
:
236 case BPF_MISC
|BPF_TAX
:
239 case BPF_MISC
|BPF_TXA
:
258 * Handle ancillary data, which are impossible
259 * (or very difficult) to get parsing packet contents.
261 switch (k
-SKF_AD_OFF
) {
262 case SKF_AD_PROTOCOL
:
263 A
= ntohs(skb
->protocol
);
269 A
= skb
->dev
->ifindex
;
280 * sk_chk_filter - verify socket filter code
281 * @filter: filter to verify
282 * @flen: length of filter
284 * Check the user's filter code. If we let some ugly
285 * filter code slip through kaboom! The filter must contain
286 * no references or jumps that are out of range, no illegal
287 * instructions, and must end with a RET instruction.
289 * All jumps are forward as they are not signed.
291 * Returns 0 if the rule set is legal or -EINVAL if not.
293 int sk_chk_filter(struct sock_filter
*filter
, int flen
)
295 struct sock_filter
*ftest
;
298 if (flen
== 0 || flen
> BPF_MAXINSNS
)
301 /* check the filter code now */
302 for (pc
= 0; pc
< flen
; pc
++) {
305 /* Only allow valid instructions */
306 switch (ftest
->code
) {
307 case BPF_ALU
|BPF_ADD
|BPF_K
:
308 case BPF_ALU
|BPF_ADD
|BPF_X
:
309 case BPF_ALU
|BPF_SUB
|BPF_K
:
310 case BPF_ALU
|BPF_SUB
|BPF_X
:
311 case BPF_ALU
|BPF_MUL
|BPF_K
:
312 case BPF_ALU
|BPF_MUL
|BPF_X
:
313 case BPF_ALU
|BPF_DIV
|BPF_X
:
314 case BPF_ALU
|BPF_AND
|BPF_K
:
315 case BPF_ALU
|BPF_AND
|BPF_X
:
316 case BPF_ALU
|BPF_OR
|BPF_K
:
317 case BPF_ALU
|BPF_OR
|BPF_X
:
318 case BPF_ALU
|BPF_LSH
|BPF_K
:
319 case BPF_ALU
|BPF_LSH
|BPF_X
:
320 case BPF_ALU
|BPF_RSH
|BPF_K
:
321 case BPF_ALU
|BPF_RSH
|BPF_X
:
322 case BPF_ALU
|BPF_NEG
:
323 case BPF_LD
|BPF_W
|BPF_ABS
:
324 case BPF_LD
|BPF_H
|BPF_ABS
:
325 case BPF_LD
|BPF_B
|BPF_ABS
:
326 case BPF_LD
|BPF_W
|BPF_LEN
:
327 case BPF_LD
|BPF_W
|BPF_IND
:
328 case BPF_LD
|BPF_H
|BPF_IND
:
329 case BPF_LD
|BPF_B
|BPF_IND
:
331 case BPF_LDX
|BPF_W
|BPF_LEN
:
332 case BPF_LDX
|BPF_B
|BPF_MSH
:
333 case BPF_LDX
|BPF_IMM
:
334 case BPF_MISC
|BPF_TAX
:
335 case BPF_MISC
|BPF_TXA
:
340 /* Some instructions need special checks */
342 case BPF_ALU
|BPF_DIV
|BPF_K
:
343 /* check for division by zero */
349 case BPF_LDX
|BPF_MEM
:
352 /* check for invalid memory addresses */
353 if (ftest
->k
>= BPF_MEMWORDS
)
359 * Note, the large ftest->k might cause loops.
360 * Compare this with conditional jumps below,
361 * where offsets are limited. --ANK (981016)
363 if (ftest
->k
>= (unsigned)(flen
-pc
-1))
367 case BPF_JMP
|BPF_JEQ
|BPF_K
:
368 case BPF_JMP
|BPF_JEQ
|BPF_X
:
369 case BPF_JMP
|BPF_JGE
|BPF_K
:
370 case BPF_JMP
|BPF_JGE
|BPF_X
:
371 case BPF_JMP
|BPF_JGT
|BPF_K
:
372 case BPF_JMP
|BPF_JGT
|BPF_X
:
373 case BPF_JMP
|BPF_JSET
|BPF_K
:
374 case BPF_JMP
|BPF_JSET
|BPF_X
:
375 /* for conditionals both must be safe */
376 if (pc
+ ftest
->jt
+ 1 >= flen
||
377 pc
+ ftest
->jf
+ 1 >= flen
)
386 return (BPF_CLASS(filter
[flen
- 1].code
) == BPF_RET
) ? 0 : -EINVAL
;
390 * sk_filter_rcu_release: Release a socket filter by rcu_head
391 * @rcu: rcu_head that contains the sk_filter to free
393 static void sk_filter_rcu_release(struct rcu_head
*rcu
)
395 struct sk_filter
*fp
= container_of(rcu
, struct sk_filter
, rcu
);
397 sk_filter_release(fp
);
400 static void sk_filter_delayed_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
402 unsigned int size
= sk_filter_len(fp
);
404 atomic_sub(size
, &sk
->sk_omem_alloc
);
405 call_rcu_bh(&fp
->rcu
, sk_filter_rcu_release
);
409 * sk_attach_filter - attach a socket filter
410 * @fprog: the filter program
411 * @sk: the socket to use
413 * Attach the user's filter code. We first run some sanity checks on
414 * it to make sure it does not explode on us later. If an error
415 * occurs or there is insufficient memory for the filter a negative
416 * errno code is returned. On success the return is zero.
418 int sk_attach_filter(struct sock_fprog
*fprog
, struct sock
*sk
)
420 struct sk_filter
*fp
, *old_fp
;
421 unsigned int fsize
= sizeof(struct sock_filter
) * fprog
->len
;
424 /* Make sure new filter is there and in the right amounts. */
425 if (fprog
->filter
== NULL
)
428 fp
= sock_kmalloc(sk
, fsize
+sizeof(*fp
), GFP_KERNEL
);
431 if (copy_from_user(fp
->insns
, fprog
->filter
, fsize
)) {
432 sock_kfree_s(sk
, fp
, fsize
+sizeof(*fp
));
436 atomic_set(&fp
->refcnt
, 1);
437 fp
->len
= fprog
->len
;
439 err
= sk_chk_filter(fp
->insns
, fp
->len
);
441 sk_filter_uncharge(sk
, fp
);
446 old_fp
= rcu_dereference(sk
->sk_filter
);
447 rcu_assign_pointer(sk
->sk_filter
, fp
);
448 rcu_read_unlock_bh();
451 sk_filter_delayed_uncharge(sk
, old_fp
);
455 int sk_detach_filter(struct sock
*sk
)
458 struct sk_filter
*filter
;
461 filter
= rcu_dereference(sk
->sk_filter
);
463 rcu_assign_pointer(sk
->sk_filter
, NULL
);
464 sk_filter_delayed_uncharge(sk
, filter
);
467 rcu_read_unlock_bh();
471 EXPORT_SYMBOL(sk_chk_filter
);
472 EXPORT_SYMBOL(sk_run_filter
);