Is fork() hook ever possible?

[Andrey Chernov]

On Mon, Nov 14, 2011 at 06:08:55PM -0500, David Schultz wrote:
> Not quite.  OpenBSD's implementation is more careful.  I just
> noticed a funny thing about FreeBSD's KERN_ARND sysctl: If the
> random device isn't (or can't be) loaded, KERN_ARND silently
> decides to initialize itself with the output of random().  This
> means that whatever minuscule amount of entropy it might have
> picked up from the clock is reduced to a maximum of 31 bits.
> That's a fantastic way to provide a false sense of security...

I agree.

Lets separate two things: "no /dev/random for jails" and "no random kernel 
module is loaded".
IMHO kernel module should _not_ be optional anymore, it solves problem you 
describe and all similar problems at once. 

Adding KERN_ARND to arc4random() at this moment solves "no /dev/random for 
jails" problem alone and _not_ pretends to solve "no random kernel module 
is loaded" problem. When random kernel module will become non-optional, 
KERN_ARND automagically makes good security in that place too.

P.S. Do I answer your doubts about &rdat key initialization in my prev. 
posting?

-- 
http://ache.vniz.net/