Re: [Buildroot] [PATCH v2] package/urandom-scripts: hash old seed with new seed when saving

[David Laight <David.Laight@ACULAB.COM>]

From: Yann E. MORIN
> Sent: 24 March 2022 10:09
> 
> David, All,
> 
> On 2022-03-24 09:15 +0000, David Laight spake thusly:
> > From: Yann E. MORIN
> > > Sent: 24 March 2022 08:25
> > >
> > > +Peter: candidate for backporting as a security fix
> > Probably not - the security fix is the code that actually
> > initialises the RNG.
> 
> Not sure I understood... As Jason explained, we have so far been saving
> a seed from an RNG that is probably partially deterministic; that is the
> security issue.

I'm pretty sure the seed is also saved during shutdown.
So a normal startup doesn't rely on the seed saved by the startup script.
(Although is would be better to use a background script
to save after (say) 30 minutes of operation.)

> The way we are seeding the RNG fundamentally does not change, because
> we are not crediting any entropy with whatever we put in there. So,
> whether we write something, anything, or nothing at all, has no impact
> on the strength of the RNG.

Right - the problem is that RNG isn't being given any entropy
at boot time.
So it is very weak.
The only point in saving/loading a seed is to give the RNG some
entropy at boot time.
Otherwise programs run from startup scripts that read the RNG
to get session keys don't get very good values at all.
That is the real security problem.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot