Re: [PATCH] crypto: exynos-rng - Remove exynos-rng driver

[Peter Griffin]

Hi Eric,

On Wed, 10 Jun 2026 at 19:39, Eric Biggers <ebiggers@xxxxxxxxxx> wrote:
>
> On Wed, Jun 10, 2026 at 03:46:54PM +0100, Peter Griffin wrote:
> > Hi Eric,
> >
> > On Sun, 31 May 2026 at 19:02, Eric Biggers <ebiggers@xxxxxxxxxx> wrote:
> > >
> > > This driver has no purpose.  It doesn't feed into the Linux RNG, nor
> > > does it implement the hwrng interface.  It is accessible only via the
> > > "rng" algorithm type of AF_ALG, which isn't used in practice.  Everyone
> > > uses either the Linux RNG, or rarely /dev/hwrng.
> > >
> > > Moreover, this is a PRNG whose only source of entropy is the 160-bit
> > > seed the user passes in.  So this can be used only by a user who already
> > > has a source of cryptographically secure random numbers, such as
> > > /dev/random.  Which they can, and do, just use in the first place.
> > >
> > > Just remove this driver.  There's no need to keep useless code around.
> > >
> > > Note that the other crypto_rng drivers in drivers/crypto/ are similarly
> > > unused and are being removed too.  This commit just handles exynos-rng.
> > >
> > > Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx>
> > > ---
> >
> > If the driver is being removed, should the binding documentation for
> > this driver not also be deleted (see
> > Documentation/devicetree/bindings/rng/samsung,exynos4-rng.yaml)?
> >
> > Peter
>
> In other discussions I've been told that devicetree bindings are
> hardware descriptions that should still exist even if there is no
> driver.  It doesn't make a lot of sense, but it seems to be what the
> devicetree people want.  I expect there would be objections to removing
> this binding.

Ok thanks for confirming.

Peter.