On Thu, Feb 28, 2019 at 03:51:41PM +0800, Herbert Xu wrote:
> On Thu, Feb 28, 2019 at 10:04:49AM +0300, Vitaly Chikunov wrote:
> >
> > It seems that you insist on set_params to be removed and both key and
> > params to be passed into set_{pub,priv}_key. This means reworking all
> > existing RSA drivers and callers, right? Can you please confirm that
> > huge rework to avoid misunderstanding?
>
> I don't understand why we even need to touch the existing RSA
> drivers. Nothing needs to change as far as they're concerned.
>
> Only the new algorithms would need to decode the extra parameters
> in the key stream.
int (*set_pub_key)(struct crypto_akcipher *tfm, const void *key,
unsigned int keylen);
int (*set_priv_key)(struct crypto_akcipher *tfm, const void *key,
unsigned int keylen);
So you want `keylen' not to cover parameters data, but parameters
actually present in key after `keylen' bytes (in come packed format)?
(And if there is no parameters appended, there is still appended some
marker, like 0, to signify that there is no parameters.)
This looks a bit counter-intuitive usage of arguments (as argument
signifying length does not cover all arguments data), is this ok to you?
More intuitive would be to add at least paramlen argument to signify how
much data is appended. Or (if we add argument anyway) additional
const void *params, unsigned int paramlen - which callers who don't
have params will pass NULL, and RSA drivers just ignore.
