Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: ATF tests still failing to complete within 2 hours on amd64



On Fri, 2 Dec 2011 15:52:50 -0500
Thor Lancelot Simon <tls%panix.com@localhost> wrote:

> > Unrelated to this issue and only out of curiosity: Since rump instances
> > are disconnected from the kernel entropy, are they still able to start
> > with a decent entropy?
> 
> What is this "still"?  Before, randomness (so-called) in rump was provided
> by keying RC4 with stack junk at rump process startup.
> 
> Now, it's as good as your libc arc4random() is, which is probably much
> better.

Perhaps I should reformulate: can rump take advantage of the fact that
the system is already running, in order to start with better entropy
than is normally available at reboot?  Not that I claim that it
previously could on netbsd-5 (I have little experience with rump; Pooka
told me that it wasn't supported on netbsd-5 when I filed (unrelated)
PRs).

If I understand your answer, rump instances can indeed take advantage
of existing entropy in the host, as the host libc strong prng is used
to seed the new rump instances prng state.  I think that this is fine.

Thanks again,
-- 
Matt


Home | Main Index | Thread Index | Old Index