[Freedombox-discuss] Entropy from the soundcard
sandyinchina at gmail.com
Fri Sep 16 01:06:42 UTC 2011
On Thu, Sep 15, 2011 at 8:13 PM, Michael Rogers <m-- at gmx.com> wrote:
> Thanks, I've had a quick look at the source and it looks promising.
> Unlike Turbid it doesn't try to produce high-quality randomness on its
> own, it just contributes bits to the kernel's entropy pool.
Turbid can do that too, of course.
> So I guess
> the next question is, what happens if you contribute completely
> predictable bits to the entropy pool? Would it harm the quality of
> /dev/random to do so?
No. All the mixing operations in the driver are invertible. They
never do x = input, always x ^= input or x += input, so adding
known bits does not destroy entropy but only moves it around.
Writes of known data to /dev/random are harmless.
What can do damage is increasing the /dev/random entropy
estimate by more than the number of bits you added. That
cannot be done by writing. Increasing the estimate is done
with a root--only ioctl call.
More information about the Freedombox-discuss