Re: [PATCH] filter: Optimize instruction revalidation code.

[Changli Gao <xiaosuo@gmail.com>]

On Wed, Nov 17, 2010 at 6:13 AM, Hagen Paul Pfeifer <hagen@jauu.net> wrote:
> * Tetsuo Handa | 2010-11-16 22:08:50 [+0900]:
>
>>--- a/net/core/filter.c
>>+++ b/net/core/filter.c
>>@@ -383,7 +383,57 @@ EXPORT_SYMBOL(sk_run_filter);
>>  */
>> int sk_chk_filter(struct sock_filter *filter, int flen)
>> {
>>-      struct sock_filter *ftest;
>>+      /*
>>+       * Valid instructions are initialized to non-0.
>>+       * Invalid instructions are initialized to 0.
>>+       */
>>+      static u16 codes[] = {
>>+              [BPF_ALU|BPF_ADD|BPF_K]  = BPF_S_ALU_ADD_K + 1,
>>+              [BPF_ALU|BPF_ADD|BPF_X]  = BPF_S_ALU_ADD_X + 1,
>>+              [BPF_ALU|BPF_SUB|BPF_K]  = BPF_S_ALU_SUB_K + 1,
> [...]
>
> Maybe I don't get it, but you increment the opcode by one, but you never
> increment the opcode in sk_run_filter() - do I miss something? Did you test
> the your patch (a trivial tcpdump rule should be sufficient)?
>

+               code = codes[code];
+               if (!code--)
+                       return -EINVAL;

But how about this:

enum {
        BPF_S_RET_K = 1,

-- 
Regards,
Changli Gao(xiaosuo@gmail.com)