Re: [PATCH] (0/4) Entropy accounting fixes

[Kent Borg <kentborg@borg.org>]

On Mon, Sep 09, 2002 at 02:57:34PM -0500, Oliver Xymoron wrote:
> > Why not use the power-on state of DRAM as a source of initial pool
> > entropy?  
> 
> There's actually not a lot there, and what is there is not random, but
> rather a rapidly fading "ghost" of what was there last. 

Which means it is not completely predictable, which means there is
going to be some entropy in there...

> And for most folks, this gets wiped by POST or the kernel long
> before the RNG gets its hands on it.

Well now, *that's* a pain.

But for us embedded folks who control our boot ROM there is the
potential to grab some entropy from that melting ghost.
  
> Nonetheless, there's no reason not to take whatever state we get
> when we allocate our pool. Amusingly, the current code needlessly
> zeroes out its pool before using it - twice! I've already fixed this
> in my patches.

Perfect!  So if some embedded-types do manage to dig up some entropy
from the end/beginning of the world, we will have a place to put it
where it will be put to use.  Cool.


Thanks,

-kb, the Kent who suggests a comment left in that code might prevent
someone from "fixing" it at a later date.