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.
18 #include <linux/module.h>
19 #include <linux/types.h>
20 #include <linux/sched.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 <linux/filter.h>
38 /* No hurry in this branch */
39 static void *__load_pointer(struct sk_buff *skb, int k)
44 ptr = skb->nh.raw + k - SKF_NET_OFF;
45 else if (k >= SKF_LL_OFF)
46 ptr = skb->mac.raw + k - SKF_LL_OFF;
48 if (ptr >= skb->head && ptr < skb->tail)
53 static inline void *load_pointer(struct sk_buff *skb, int k,
54 unsigned int size, void *buffer)
57 return skb_header_pointer(skb, k, size, buffer);
61 return __load_pointer(skb, k);
66 * sk_run_filter - run a filter on a socket
67 * @skb: buffer to run the filter on
68 * @filter: filter to apply
69 * @flen: length of filter
71 * Decode and apply filter instructions to the skb->data.
72 * Return length to keep, 0 for none. skb is the data we are
73 * filtering, filter is the array of filter instructions, and
74 * len is the number of filter blocks in the array.
77 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(*(u32 *)ptr);
184 case BPF_LD|BPF_H|BPF_ABS:
187 ptr = load_pointer(skb, k, 2, &tmp);
189 A = ntohs(*(u16 *)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:
243 return ((unsigned int)fentry->k);
245 return ((unsigned int)A);
253 /* Invalid instruction counts as RET */
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 = htons(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 instructions
287 * and no backward jumps. It must end with a RET instruction
289 * Returns 0 if the rule set is legal or a negative errno code if not.
291 int sk_chk_filter(struct sock_filter *filter, int flen)
293 struct sock_filter *ftest;
296 if (((unsigned int)flen >= (~0U / sizeof(struct sock_filter))) || flen == 0)
299 /* check the filter code now */
300 for (pc = 0; pc < flen; pc++) {
301 /* all jumps are forward as they are not signed */
303 if (BPF_CLASS(ftest->code) == BPF_JMP) {
304 /* but they mustn't jump off the end */
305 if (BPF_OP(ftest->code) == BPF_JA) {
307 * Note, the large ftest->k might cause loops.
308 * Compare this with conditional jumps below,
309 * where offsets are limited. --ANK (981016)
311 if (ftest->k >= (unsigned)(flen-pc-1))
314 /* for conditionals both must be safe */
315 if (pc + ftest->jt +1 >= flen ||
316 pc + ftest->jf +1 >= flen)
321 /* check for division by zero -Kris Katterjohn 2005-10-30 */
322 if (ftest->code == (BPF_ALU|BPF_DIV|BPF_K) && ftest->k == 0)
325 /* check that memory operations use valid addresses. */
326 if (ftest->k >= BPF_MEMWORDS) {
327 /* but it might not be a memory operation... */
328 switch (ftest->code) {
332 case BPF_LDX|BPF_MEM:
339 * The program must end with a return. We don't care where they
340 * jumped within the script (its always forwards) but in the end
341 * they _will_ hit this.
343 return (BPF_CLASS(filter[flen - 1].code) == BPF_RET) ? 0 : -EINVAL;
347 * sk_attach_filter - attach a socket filter
348 * @fprog: the filter program
349 * @sk: the socket to use
351 * Attach the user's filter code. We first run some sanity checks on
352 * it to make sure it does not explode on us later. If an error
353 * occurs or there is insufficient memory for the filter a negative
354 * errno code is returned. On success the return is zero.
356 int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk)
358 struct sk_filter *fp;
359 unsigned int fsize = sizeof(struct sock_filter) * fprog->len;
362 /* Make sure new filter is there and in the right amounts. */
363 if (fprog->filter == NULL || fprog->len > BPF_MAXINSNS)
366 fp = sock_kmalloc(sk, fsize+sizeof(*fp), GFP_KERNEL);
369 if (copy_from_user(fp->insns, fprog->filter, fsize)) {
370 sock_kfree_s(sk, fp, fsize+sizeof(*fp));
374 atomic_set(&fp->refcnt, 1);
375 fp->len = fprog->len;
377 err = sk_chk_filter(fp->insns, fp->len);
379 struct sk_filter *old_fp;
381 spin_lock_bh(&sk->sk_lock.slock);
382 old_fp = sk->sk_filter;
384 spin_unlock_bh(&sk->sk_lock.slock);
389 sk_filter_release(sk, fp);
393 EXPORT_SYMBOL(sk_chk_filter);
394 EXPORT_SYMBOL(sk_run_filter);