]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/core/filter.c
[NET]: Use newer is_multicast_ether_addr() in some files
[net-next-2.6.git] / net / core / filter.c
CommitLineData
1da177e4
LT
1/*
2 * Linux Socket Filter - Kernel level socket filtering
3 *
4 * Author:
5 * Jay Schulist <jschlst@samba.org>
6 *
7 * Based on the design of:
8 * - The Berkeley Packet Filter
9 *
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.
14 *
15 * Andi Kleen - Fix a few bad bugs and races.
93699863 16 * Kris Katterjohn - Added many additional checks in sk_chk_filter()
1da177e4
LT
17 */
18
19#include <linux/module.h>
20#include <linux/types.h>
21#include <linux/sched.h>
22#include <linux/mm.h>
23#include <linux/fcntl.h>
24#include <linux/socket.h>
25#include <linux/in.h>
26#include <linux/inet.h>
27#include <linux/netdevice.h>
28#include <linux/if_packet.h>
29#include <net/ip.h>
30#include <net/protocol.h>
31#include <linux/skbuff.h>
32#include <net/sock.h>
33#include <linux/errno.h>
34#include <linux/timer.h>
35#include <asm/system.h>
36#include <asm/uaccess.h>
37#include <linux/filter.h>
38
39/* No hurry in this branch */
0b05b2a4 40static void *__load_pointer(struct sk_buff *skb, int k)
1da177e4
LT
41{
42 u8 *ptr = NULL;
43
44 if (k >= SKF_NET_OFF)
45 ptr = skb->nh.raw + k - SKF_NET_OFF;
46 else if (k >= SKF_LL_OFF)
47 ptr = skb->mac.raw + k - SKF_LL_OFF;
48
49 if (ptr >= skb->head && ptr < skb->tail)
50 return ptr;
51 return NULL;
52}
53
0b05b2a4
PM
54static inline void *load_pointer(struct sk_buff *skb, int k,
55 unsigned int size, void *buffer)
56{
57 if (k >= 0)
58 return skb_header_pointer(skb, k, size, buffer);
59 else {
60 if (k >= SKF_AD_OFF)
61 return NULL;
62 return __load_pointer(skb, k);
63 }
64}
65
1da177e4
LT
66/**
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
71 *
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.
76 */
77
78int sk_run_filter(struct sk_buff *skb, struct sock_filter *filter, int flen)
79{
1da177e4 80 struct sock_filter *fentry; /* We walk down these */
0b05b2a4 81 void *ptr;
1da177e4
LT
82 u32 A = 0; /* Accumulator */
83 u32 X = 0; /* Index Register */
84 u32 mem[BPF_MEMWORDS]; /* Scratch Memory Store */
0b05b2a4 85 u32 tmp;
1da177e4
LT
86 int k;
87 int pc;
88
89 /*
90 * Process array of filter instructions.
91 */
92 for (pc = 0; pc < flen; pc++) {
93 fentry = &filter[pc];
94
95 switch (fentry->code) {
96 case BPF_ALU|BPF_ADD|BPF_X:
97 A += X;
98 continue;
99 case BPF_ALU|BPF_ADD|BPF_K:
100 A += fentry->k;
101 continue;
102 case BPF_ALU|BPF_SUB|BPF_X:
103 A -= X;
104 continue;
105 case BPF_ALU|BPF_SUB|BPF_K:
106 A -= fentry->k;
107 continue;
108 case BPF_ALU|BPF_MUL|BPF_X:
109 A *= X;
110 continue;
111 case BPF_ALU|BPF_MUL|BPF_K:
112 A *= fentry->k;
113 continue;
114 case BPF_ALU|BPF_DIV|BPF_X:
115 if (X == 0)
116 return 0;
117 A /= X;
118 continue;
119 case BPF_ALU|BPF_DIV|BPF_K:
1da177e4
LT
120 A /= fentry->k;
121 continue;
122 case BPF_ALU|BPF_AND|BPF_X:
123 A &= X;
124 continue;
125 case BPF_ALU|BPF_AND|BPF_K:
126 A &= fentry->k;
127 continue;
128 case BPF_ALU|BPF_OR|BPF_X:
129 A |= X;
130 continue;
131 case BPF_ALU|BPF_OR|BPF_K:
132 A |= fentry->k;
133 continue;
134 case BPF_ALU|BPF_LSH|BPF_X:
135 A <<= X;
136 continue;
137 case BPF_ALU|BPF_LSH|BPF_K:
138 A <<= fentry->k;
139 continue;
140 case BPF_ALU|BPF_RSH|BPF_X:
141 A >>= X;
142 continue;
143 case BPF_ALU|BPF_RSH|BPF_K:
144 A >>= fentry->k;
145 continue;
146 case BPF_ALU|BPF_NEG:
147 A = -A;
148 continue;
149 case BPF_JMP|BPF_JA:
150 pc += fentry->k;
151 continue;
152 case BPF_JMP|BPF_JGT|BPF_K:
153 pc += (A > fentry->k) ? fentry->jt : fentry->jf;
154 continue;
155 case BPF_JMP|BPF_JGE|BPF_K:
156 pc += (A >= fentry->k) ? fentry->jt : fentry->jf;
157 continue;
158 case BPF_JMP|BPF_JEQ|BPF_K:
159 pc += (A == fentry->k) ? fentry->jt : fentry->jf;
160 continue;
161 case BPF_JMP|BPF_JSET|BPF_K:
162 pc += (A & fentry->k) ? fentry->jt : fentry->jf;
163 continue;
164 case BPF_JMP|BPF_JGT|BPF_X:
165 pc += (A > X) ? fentry->jt : fentry->jf;
166 continue;
167 case BPF_JMP|BPF_JGE|BPF_X:
168 pc += (A >= X) ? fentry->jt : fentry->jf;
169 continue;
170 case BPF_JMP|BPF_JEQ|BPF_X:
171 pc += (A == X) ? fentry->jt : fentry->jf;
172 continue;
173 case BPF_JMP|BPF_JSET|BPF_X:
174 pc += (A & X) ? fentry->jt : fentry->jf;
175 continue;
176 case BPF_LD|BPF_W|BPF_ABS:
177 k = fentry->k;
178 load_w:
0b05b2a4
PM
179 ptr = load_pointer(skb, k, 4, &tmp);
180 if (ptr != NULL) {
181 A = ntohl(*(u32 *)ptr);
182 continue;
1da177e4 183 }
1198ad00 184 break;
1da177e4
LT
185 case BPF_LD|BPF_H|BPF_ABS:
186 k = fentry->k;
187 load_h:
0b05b2a4
PM
188 ptr = load_pointer(skb, k, 2, &tmp);
189 if (ptr != NULL) {
190 A = ntohs(*(u16 *)ptr);
191 continue;
1da177e4 192 }
1198ad00 193 break;
1da177e4
LT
194 case BPF_LD|BPF_B|BPF_ABS:
195 k = fentry->k;
196load_b:
0b05b2a4
PM
197 ptr = load_pointer(skb, k, 1, &tmp);
198 if (ptr != NULL) {
199 A = *(u8 *)ptr;
200 continue;
1da177e4 201 }
1198ad00 202 break;
1da177e4 203 case BPF_LD|BPF_W|BPF_LEN:
3154e540 204 A = skb->len;
1da177e4
LT
205 continue;
206 case BPF_LDX|BPF_W|BPF_LEN:
3154e540 207 X = skb->len;
1da177e4
LT
208 continue;
209 case BPF_LD|BPF_W|BPF_IND:
210 k = X + fentry->k;
211 goto load_w;
212 case BPF_LD|BPF_H|BPF_IND:
213 k = X + fentry->k;
214 goto load_h;
215 case BPF_LD|BPF_B|BPF_IND:
216 k = X + fentry->k;
217 goto load_b;
218 case BPF_LDX|BPF_B|BPF_MSH:
0b05b2a4
PM
219 ptr = load_pointer(skb, fentry->k, 1, &tmp);
220 if (ptr != NULL) {
221 X = (*(u8 *)ptr & 0xf) << 2;
222 continue;
223 }
224 return 0;
1da177e4
LT
225 case BPF_LD|BPF_IMM:
226 A = fentry->k;
227 continue;
228 case BPF_LDX|BPF_IMM:
229 X = fentry->k;
230 continue;
231 case BPF_LD|BPF_MEM:
232 A = mem[fentry->k];
233 continue;
234 case BPF_LDX|BPF_MEM:
235 X = mem[fentry->k];
236 continue;
237 case BPF_MISC|BPF_TAX:
238 X = A;
239 continue;
240 case BPF_MISC|BPF_TXA:
241 A = X;
242 continue;
243 case BPF_RET|BPF_K:
244 return ((unsigned int)fentry->k);
245 case BPF_RET|BPF_A:
246 return ((unsigned int)A);
247 case BPF_ST:
248 mem[fentry->k] = A;
249 continue;
250 case BPF_STX:
251 mem[fentry->k] = X;
252 continue;
253 default:
93699863 254 WARN_ON(1);
1da177e4
LT
255 return 0;
256 }
257
258 /*
259 * Handle ancillary data, which are impossible
260 * (or very difficult) to get parsing packet contents.
261 */
262 switch (k-SKF_AD_OFF) {
263 case SKF_AD_PROTOCOL:
264 A = htons(skb->protocol);
265 continue;
266 case SKF_AD_PKTTYPE:
267 A = skb->pkt_type;
268 continue;
269 case SKF_AD_IFINDEX:
270 A = skb->dev->ifindex;
271 continue;
272 default:
273 return 0;
274 }
275 }
276
277 return 0;
278}
279
280/**
281 * sk_chk_filter - verify socket filter code
282 * @filter: filter to verify
283 * @flen: length of filter
284 *
285 * Check the user's filter code. If we let some ugly
286 * filter code slip through kaboom! The filter must contain
93699863
KK
287 * no references or jumps that are out of range, no illegal
288 * instructions, and must end with a RET instruction.
1da177e4
LT
289 *
290 * Returns 0 if the rule set is legal or a negative errno code if not.
291 */
292int sk_chk_filter(struct sock_filter *filter, int flen)
293{
294 struct sock_filter *ftest;
295 int pc;
296
1b93ae64 297 if (flen == 0 || flen > BPF_MAXINSNS)
1da177e4
LT
298 return -EINVAL;
299
300 /* check the filter code now */
301 for (pc = 0; pc < flen; pc++) {
302 /* all jumps are forward as they are not signed */
303 ftest = &filter[pc];
1da177e4 304
93699863
KK
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:
330 case BPF_LD|BPF_IMM:
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:
336 case BPF_RET|BPF_K:
337 case BPF_RET|BPF_A:
338 break;
339
340 /* Some instructions need special checks */
fb0d366b 341
93699863
KK
342 case BPF_ALU|BPF_DIV|BPF_K:
343 /* check for division by zero */
344 if (ftest->k == 0)
1da177e4 345 return -EINVAL;
93699863
KK
346 break;
347
348 case BPF_LD|BPF_MEM:
349 case BPF_LDX|BPF_MEM:
350 case BPF_ST:
351 case BPF_STX:
352 /* check for invalid memory addresses */
353 if (ftest->k >= BPF_MEMWORDS)
354 return -EINVAL;
355 break;
356
357 case BPF_JMP|BPF_JA:
358 /*
359 * Note, the large ftest->k might cause loops.
360 * Compare this with conditional jumps below,
361 * where offsets are limited. --ANK (981016)
362 */
363 if (ftest->k >= (unsigned)(flen-pc-1))
364 return -EINVAL;
365 break;
366
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)
378 return -EINVAL;
379 break;
380
381 default:
382 return -EINVAL;
1da177e4
LT
383 }
384 }
385
386 /*
387 * The program must end with a return. We don't care where they
388 * jumped within the script (its always forwards) but in the end
389 * they _will_ hit this.
390 */
391 return (BPF_CLASS(filter[flen - 1].code) == BPF_RET) ? 0 : -EINVAL;
392}
393
394/**
395 * sk_attach_filter - attach a socket filter
396 * @fprog: the filter program
397 * @sk: the socket to use
398 *
399 * Attach the user's filter code. We first run some sanity checks on
400 * it to make sure it does not explode on us later. If an error
401 * occurs or there is insufficient memory for the filter a negative
402 * errno code is returned. On success the return is zero.
403 */
404int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk)
405{
406 struct sk_filter *fp;
407 unsigned int fsize = sizeof(struct sock_filter) * fprog->len;
408 int err;
409
410 /* Make sure new filter is there and in the right amounts. */
1b93ae64 411 if (fprog->filter == NULL)
1da177e4
LT
412 return -EINVAL;
413
414 fp = sock_kmalloc(sk, fsize+sizeof(*fp), GFP_KERNEL);
415 if (!fp)
416 return -ENOMEM;
417 if (copy_from_user(fp->insns, fprog->filter, fsize)) {
418 sock_kfree_s(sk, fp, fsize+sizeof(*fp));
419 return -EFAULT;
420 }
421
422 atomic_set(&fp->refcnt, 1);
423 fp->len = fprog->len;
424
425 err = sk_chk_filter(fp->insns, fp->len);
426 if (!err) {
427 struct sk_filter *old_fp;
428
429 spin_lock_bh(&sk->sk_lock.slock);
430 old_fp = sk->sk_filter;
431 sk->sk_filter = fp;
432 spin_unlock_bh(&sk->sk_lock.slock);
433 fp = old_fp;
434 }
435
436 if (fp)
437 sk_filter_release(sk, fp);
438 return err;
439}
440
441EXPORT_SYMBOL(sk_chk_filter);
442EXPORT_SYMBOL(sk_run_filter);