> From: Theo de Raadt <dera...@cvs.openbsd.org> > Subject: Re: /dev/srandom vs. /dev/arandom > To: "James Hozier" <guitars...@yahoo.com> > Cc: misc@openbsd.org > Date: Tuesday, October 18, 2011, 12:53 AM > > I heard that since 4.9, there > has been some changes to the > > /dev/randoms in OpenBSD. I'm unsure of what the > changes exactly are, > > but for confidentiality in terms of entire hard drives > (talking > > terabytes of SATAII hard drives), would /dev/srandom > still be the best > > suitable for this task? > > There is now only one random device, /dev/random. > > It is all PRNG, but our PRNG is very good. The pool > management has a > set of nested data recursions that mix newly collected > randomness > (from interrupts and such) with the timing of extractions, > of course > at the same time that all entropy requests are being > segmented in > invisible ways amongst many consumers (especially those > small requests > made so often by our kernel code). It acted like that > before, but it > is now even better. > > And yes, it is a lot faster. > > > Last I remember, /dev/arandom was much too > > slow since I could not do enough on my computer to > create enough > > entropy to randomize my disks before an entire year > passed, heh > > I'm seeing 150MB/sec of output on a fairly fast machine. > > It was slower before, but not that much slower. I > think you are > exaggerating. > >
Perhaps I was confusing my memory with a certain other operating system's /dev/random. In any case, I'm getting just under 600KB/s on average with /dev/random. This is on a rather old machine, so I guess it's not too bad.
