Re: interrupts & pmf/legacy shutdown hooks (was Re: CVS commit:src/sys/arch)

[Simon Burge <simonb%NetBSD.org@localhost>]

Matthias Drochner wrote:

> dyoung%pobox.com@localhost said:
> > following ad's finding that disabling interrupts causes panics
> 
> Sorry, "disabling interrupts causes panics" is a little too
> little justification for such a change which has consequences
> for a lot of complex drivers. Interrupts were disabled at that
> point ~forever. And if something suddenly starts to cause panics
> (I didn't succeed in finding a report in the mailing lists yet),
> then certainly a particular driver is to blame.
> And even if there was a feeling that enabling interrupts on
> shutdown would be a step into the right direction, it would
> have to be done for all ports, so that mi drivers can count
> on a well-defined environment.
> 
> > Please, back out the change
> 
> I'd suggest to leave things as they are now until we get
> an indication thst enabling interrupts actually fixes
> anything, and that having the ports behave inconsistently
> is the lesser evil.

This breaks kern/38655 again (ld on twa/twe), which had been working
for a while.

The fix as suggested in the PR is simple and tested, I can commit it
at a moment's notice (it's still in my -current tree).  It'd just be
nice to decide once and for all what the best way is instead of flip
flopping around on this.

Simon.