random(4) plugin infrastructure for mulitple RNG in a modular fashion

[Scott Long]

On Aug 7, 2013, at 11:35 AM, David O'Brien <obrien at FreeBSD.org> wrote:

> Replying to an earlier comments made:
> 
> On Mon, Aug 05, 2013 at 09:42:51PM -0600, Scott Long wrote:
>>> OK - so we are agreed that obrien@'s commit didn't break anything?
>> 
>> It absolutely broke something.  It rendered my remote test machines useless
>> after what should have been a routine source upgrade.  Even figuring
>> out what happened and mashing on the remote keyboard only succeeded in
>> making the machines limp forward with no /dev/random device.
> 
> Hi Scott,
> Please give more detail about the issue if you can.  I would really like
> to understand how the changeset negatively impacted you in order to fix
> the issue.
> 

We only compile and deploy a whitelist subset of modules, so the PRNG
modules were not compiled into the kernel nor present in /boot.  When the
machine came up, it paused waiting for keyboard input.  In one iteration, it
released the hold after about 100 characters of quick typing.  In another
iteration, it waited as I slowly typed only a few characters a minute, until I
gave up.  In a third iteration, it waited indefinitely as I typed nothing, and I
gave up after 5-10 minutes.

In the first iteration, even after the system continued, various things complained
and/or failed with a complaint of there being no /dev/random.  named was one
thing, but I know there were others.  We've since moved on and picked up
the post-revert sources.

Scott