On Fri, 2021-04-09 at 16:08 +0800, Hangbin Liu wrote:
> On Fri, Apr 09, 2021 at 09:08:20AM +0200, Stephan Mueller wrote:
> > > > > > > And how do you handle all the other places in the kernel that use
> > > > > > > ChaCha20 and
> > > > > > > SipHash? For example, drivers/char/random.c?
> > > > > >
> > > > > > Good question, I will check it and reply to you later.
> > > > >
> > > > > I just read the code. The drivers/char/random.c do has some fips
> > > > > specific
> > > > > parts(seems not related to crypto). After commit e192be9d9a30 ("random:
> > > > > replace
> > > > > non-blocking pool with a Chacha20-based CRNG") we moved part of chacha
> > > > > code to
> > > > > lib/chacha20.c and make that code out of control.
> > > > >
> > > > So you are saying that you removed drivers/char/random.c and
> > > > lib/chacha20.c from
> > > > your FIPS module boundary? Why not do the same for WireGuard?
> > >
> > > No, I mean this looks like a bug (using not allowed crypto in FIPS mode) and
> > > we should fix it.
> >
> > The entirety of random.c is not compliant to FIPS rules. ChaCha20 is the least
> > of the problems. SP800-90B is the challenge. This is one of the motivation of
> > the design and architecture of the LRNG allowing different types of crypto and
> > have a different approach to post-process the data.
> >
> > https://github.com/smuellerDD/lrng
>
> Thanks Stephan for this info. After offline discussion with Herbert, here is
> what he said:
>
> """
> This is not a problem in RHEL8 because the Crypto API RNG replaces /dev/random
> in FIPS mode.
> """
>
> I'm not familiar with this code, not sure how upstream handle this.
It is an open problem upstream.
Simo.
--
Simo Sorce
RHEL Crypto Team
Red Hat, Inc
