On Thu, 2006-07-06 at 15:45 +0200, Martin Langer wrote:

> +       tfm = crypto_alloc_tfm("md5", 0);
> +       if (tfm == NULL)
> +               return -1;

That's all fine and dandy, but how about including a Kconfig option
instead that makes it rely on the md5sum code?

> +       crypto_digest_init(tfm);
> +       crypto_digest_update(tfm, sg, 1);
> +       crypto_digest_final(tfm, signature);
> +       crypto_free_tfm(tfm);

None of these functions can return an error? 

I also don't like your handling of the struct scatterlist. IMHO you
should only declare one
        struct scatterlist sg;

and use &sg, instead of sg_set_buf you should use sg_init_one, and
change the following:
        crypto_digest_update(tfm, sg, 1);
to
        crypto_digest_update(tfm, &sg, 1);

johannes

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to