On Sat, Feb 2, 2013 at 2:19 AM, Paul Wouters <paul(a)nohats.ca> wrote:
On Fri, 1 Feb 2013, Matthew Garrett wrote:
> other than providing other sources of entropy, and long-term this is
> going to be fixed once everyone's moved to Ivy Bridge and has an
> unprivileged instruction to hand out entropy.
uhm I know intel really wants us to use it directly and trust them, but
we're going to run it through the kernel right? And just expose it via
/dev/random to userland yes?
... and applications will call the best-matching RNG
function from a
reputable crypto library instead of reading /dev/anything or using an
architecture-specific instruction directly, hopefully.
(That said, if you don't trust Intel to implement rdrand properly, do
you trust them not to specially recognize and "mis-execute" code
implementing the kernel /dev/random entropy pool update or other
similarly critical code? There is even that handy microcode update
mechanism that allows a hypothetical malicious Intel to adapt to
kernel code changes.)
Mirek