On Sun, Oct 06, 2002 at 02:06:23PM +0200, Martin Husemann wrote:
> It's easier to cope with a hanging controller when it worked at all - at
> autoconfig time most drivers do not take much attempts to cope with unexpected
> situations, they just bail out (leaving the device unconfigured).
> 
> Now you obviously caught it in a bad place, where it things it can handle the
> controller, but the thread initialising the scsi bus later finds out it realy
> can't. 
> 
> What happens after the command timeout? Realy nothing? That would be a serious
> bug, of course.

I suspect interrupts don't work. After the command timeout, a SCSI bus
reset is issued, which should trigger an interrupt which would get things
going again. I think this interrupt never occurs. 
I'll look at adding a timeout for SCSI bus reset.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
--