Re: Patch: rework kernel random number subsystem

On Fri, Oct 21, 2011 at 05:15:55PM -0400, Thor Lancelot Simon wrote:
> 
> WARNING:	#7 and #8 reveal some kind of synchronization or locking
> 		bug in this patch.  #8 causes the entropy pool to log to
> 		the console whenever it supplies rekeying entropy.  #7
> 		causes 'sysctl kern.urandom' to read from a cprng_strong
> 	        instance.
> 
> 		Performing around 1000 consecutive such sysctl calls will
> 		reveal corruption of the cprng_strong state: it is not
> 		rekeyed (nor should it yet be), but is corrupted in such
> 		a way that it thinks it has been, triggering the rngtest
> 		statistical test, which then fails.

The patch at http://www.panix.com/~tls/rnd2.diff addresses the correctness
issues that Christos pointed out but does *not* fix the problem described
above.

Help much appreciated.

Thor