[ovs-dev] [PATCH 3/3] random: Get random seed from /dev/urandom.
shemminger at vyatta.com
Thu Aug 12 11:40:50 PDT 2010
On Thu, 12 Aug 2010 11:28:47 -0700
Ben Pfaff <blp at nicira.com> wrote:
> On Thu, Aug 12, 2010 at 02:18:58PM -0400, Stephen Hemminger wrote:
> > On Thu, 12 Aug 2010 11:12:21 -0700
> > Ben Pfaff <blp at nicira.com> wrote:
> > > Even though this PRNG is not meant to be cryptographically secure, there is
> > > no reason not to get a high-quality seed.
> > >
> > > CC: Stephen Hemminger <shemminger at vyatta.com>
> > P.s: this is what the linux kernel uses a Tausworthe PRNG for
> > this. Its overkill, but I was worried about good distribution for
> > emulation with netem.
> Do you have concerns about the Xorshift RNG that I chose? I did read
> the original paper by George Marsaglia about it (cited in the source
> file), which convinced me that it was suitable enough for the purpose.
> I didn't look at what the kernel uses, but now I see that you're the
> author of the high-speed PRNG there, so now I understand why this commit
> caught your eye.
I think Xorshift is fine. My goal was to resolve statistical issues in
simulation (but then code got reused as it always does).
More information about the dev