Re: [PATCH v2] RANDOM: ATH9K RNG delivers zero bits of entropy

[Jason Cooper <jason@lakedaemon.net>]

Hey Ted,

On Wed, Aug 10, 2016 at 07:44:25PM -0400, Theodore Ts'o wrote:
> On Tue, Aug 09, 2016 at 02:04:44PM +0000, Jason Cooper wrote:
> > iiuc, Ted, you're saying using the hw_random framework would be
> > disasterous because despite most drivers having a default quality of 0,
> > rngd assumes 1 bit of entropy for every bit read?
> 
> Sorry, what I was trying to say (but failed) was that bypassing the
> hwrng framework and injecting entropy directly the entropy pool was
> disatrous.

Ok, whew. :)

> > Thankfully, most hw_random drivers don't set the quality.  So unless the
> > user sets the default_quality param, it's zero.
> 
> The fact that this is "most" and not "all" does scare me a little.

My recent grep showed that only virtio-rng set it to a non-zero value.

> As far as I'm concerned *all* hw_random drivers should set quality to
> zero, since it should be up to the system administrator.

Agreed.

Gathering conversation about this from a few related threads, I have one
concern.  Apparently there is some confusion in userspace consumers of
/dev/hwrng data as to the quality of it.  Specifically, rngd (spotted by
Stephan Mueller) appears to assume 1bit of entropy per 1 bit read. :-/

So, while moving ath9k-rng to the hwrng framework makes complete sense
internally, it's not so good for existing userspace assumptions.  I'd
think that timeriomem-rng falls in this same category.

In light of this, do you think it's worth the effort (I'm volunteering)
to create a subcategory of hwrng drivers that are 'environemntal' rngs?
They can contribute to the kernel entropy pools, but not to /dev/hwrng.

thx,

Jason.