On 10/27/2017 10:35 AM, Arnt Gulbrandsen wrote:

> Olaf Meeuwissen writes:
>> I have used the `haveged` package to keep my /dev/urandom "topped up"
>> when randomizing disks. Greatly shortened the time needed to fill my
>> disks. No idea about the quality of randomness, though.
>
> I looked at it now. It seems to observe some real entropy, but I think
> they overestimate the amount. Some of the events they count show up in
> some/many counters, and I don't see any attempt to estimate or account
> for covariance.
>
> So, yes, good stuff IMHO, but I don't think it's actually hundreds of
> megabits per second. Still, even 1kbps of real entropy isn't bad.
I can't imagine it being equivalent to a (non-intel/amd) hardware source
of entropy when it comes to quality of entropy - have there been any
quality analysis performed?

It is a shame IDQ is the only vendor with a PCI-e device, and also the
only vendor it seems that offers something quality and obtainable (I
can't find an open source hardware entropy device that is really for
sale right away