Date: Thu, 20 Sep 2012 09:59:17 +0100 From: Ben Laurie <benl@freebsd.org> To: Pawel Jakub Dawidek <pjd@freebsd.org> Cc: freebsd-security@freebsd.org, Jonathan Anderson <jonathan.anderson@cl.cam.ac.uk> Subject: Re: Collecting entropy from device_attach() times. Message-ID: <CAG5KPzx2oK%2BULbfbthj2XVx-%2BRtRHHbYMW1BG6Aymju-0x-vvg@mail.gmail.com> In-Reply-To: <20120919202023.GD1416@garage.freebsd.pl> References: <20120918211422.GA1400@garage.freebsd.pl> <A8FD98DD94774D00B4E5F78D3174C1B4@gmail.com> <20120919192923.GA1416@garage.freebsd.pl> <CAG5KPzyxSMZ8X4RmEhCHA=dHTLUw5mOUf-oveJtOPx8im3dpeQ@mail.gmail.com> <20120919202023.GD1416@garage.freebsd.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Sep 19, 2012 at 9:20 PM, Pawel Jakub Dawidek <pjd@freebsd.org> wrot= e: > On Wed, Sep 19, 2012 at 08:59:15PM +0100, Ben Laurie wrote: >> On Wed, Sep 19, 2012 at 8:29 PM, Pawel Jakub Dawidek <pjd@freebsd.org> w= rote: >> > On Wed, Sep 19, 2012 at 07:30:52PM +0100, Jonathan Anderson wrote: >> >> > If all the times are more or less equally probable in this range [= =85] >> >> >> >> They're very unlikely to be equally probable. It would make sense to = do some characterization of these times and their statistics: a highly non-= uniform distribution would mean that we don't actually get many bits per at= tach. >> > >> > I have times for ~2000 device_attach() calls when loading sound card >> > driver on totally idle system. If someone could take those and analyse >> > the distribution that would be great. >> > >> >> > [=85] we have more >> >> > than 19 bits of entropy from this one call, but I reduced if to fou= r >> >> > bits only, because there are devices that are much faster to attach= . >> >> > >> >> >> >> Another reason for doing the above characterization is that, if a par= ticular device_attach() really does provide 12 bits of uncertainty, it's a = shame to drop eight of them on the floor. >> > >> > Rights. That's why I've prepared another patch: >> > >> > http://people.freebsd.org/~pjd/patches/harvest_device_attach.2= .patch >> > >> > which effectively discards top ten bits, which means we expect 0.1% of >> > the attach time to be unpredictable (the attach time in most cases var= y >> > by few percent, not sure yet how much of this variation is really >> > unpredictable). >> >> This is the wrong thing to do! There's no reason to discard bits on >> input (modulo the device throwing away inputs, that is) - just reduce >> your entropy estimate. "Extra" bits do no harm. > > I 'discard' ten bits from the estimation. I don't discard them by > zeroing them out. If the number is a 26 bit value then I feed entire > number, but pass estimation of 16 bits. Sorry, should've read the code first! This is great. I also like your friend's analysis.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAG5KPzx2oK%2BULbfbthj2XVx-%2BRtRHHbYMW1BG6Aymju-0x-vvg>