2009-04-18 00:25:15

by Hartley Sweeten

Subject: Q: crypto/cipher.c

Hello all,
I have a question about the following code in crypto/cipher.c.

static void cipher_crypt_unaligned(void (*fn)(struct crypto_tfm *, u8 *,
const u8 *),
struct crypto_tfm *tfm,
u8 *dst, const u8 *src)
{
unsigned long alignmask = crypto_tfm_alg_alignmask(tfm);
unsigned int size = crypto_tfm_alg_blocksize(tfm);
u8 buffer[size + alignmask];
u8 *tmp = (u8 *)ALIGN((unsigned long)buffer, alignmask + 1);

memcpy(tmp, src, size);
fn(tfm, tmp, tmp);
memcpy(dst, tmp, size);
}

Doesn't the size of buffer need to be known at compile time for this
to actually work? Or do alignmask and size work out to be constants?

This appears to be buggy. But I may just be missing something...

Thanks,
Hartley

2009-04-20 18:41:55

by Paulo Marques

Subject: Re: Q: crypto/cipher.c

H Hartley Sweeten wrote:
> Hello all,

Hi,

> I have a question about the following code in crypto/cipher.c.
>
> static void cipher_crypt_unaligned(void (*fn)(struct crypto_tfm *, u8 *,
> const u8 *),
> struct crypto_tfm *tfm,
> u8 *dst, const u8 *src)
> {
> unsigned long alignmask = crypto_tfm_alg_alignmask(tfm);
> unsigned int size = crypto_tfm_alg_blocksize(tfm);
> u8 buffer[size + alignmask];
> u8 *tmp = (u8 *)ALIGN((unsigned long)buffer, alignmask + 1);
>
> memcpy(tmp, src, size);
> fn(tfm, tmp, tmp);
> memcpy(dst, tmp, size);
> }
>
> Doesn't the size of buffer need to be known at compile time for this
> to actually work? Or do alignmask and size work out to be constants?
>
> This appears to be buggy. But I may just be missing something...

It's a GCC extension:

http://gcc.gnu.org/onlinedocs/gcc/Variable-Length.html

--
Paulo Marques - http://www.grupopie.com

"All generalizations are false."