Re: [Qemu-devel] virtio-rng

[Jamie Lokier]

Ian Molton wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Jamie Lokier wrote:
> > Ian Molton wrote:
> >> Heres my patch to virtio-console. The device is now specified like this:
> >>
> >> - -chardev file,path=/path/to/testfile,id=test
> >> - -device virtio-console-pci,chardev=test
> 
> Note, I think the patch above is broken and slipped by me because I
> forgot to make clean - I'll doublecheck tomorrow.
> 
> > It'd be nice if some options on the qemu command line (or config file)
> > resulted in the guest kernel getting entropy (assuming a kernel with
> > the appropriate support), without having to run a special
> > entropy-injection daemon in the guest.
> 
> How, though - AFAIK linux has no drivers specifically intended to feed
> entropy into the kernels pool - all the hwrng drivers use a userspace
> tool to do so.

You're right, with the reason being hardware randomness should be
tested before being passed back to the kernel as trusted entropy, as
it might not be functioning properly or might not have the quality it
claims to have.

With VMs, in some circumstances it might be preferable to trust the
host when it says it's providing already-tested entropy.  After all
the host has total control over the guest anyway, and the host entropy
has already been run through the same checks.

So I think virtio-rng could benefit form being a special case, if the
host says "I assert this is entropy", you might inject it directly,
and thus work even with guests that aren't running the rngd daemon for
one reason or another.  (E.g. embedded system guests.)

> I still intend to submit my virtio-rng driver, if thats what you mean,
> since it presents the data via the same routes as all the other hw rng
> sources. I feel this approach has value.

Suddenly I'm intrigued by this "intend to submit" as I see a
virtio-rng driver already in the 2.6 tree... What have I misunderstood?

Thanks,
-- Jamie