Re: [patch] O(1) scheduler, -H5

[Russell King]

On Fri, Jan 11, 2002 at 12:42:14PM +0100, Andi Kleen wrote:
> When they hold the kernel lock in addition to the global cli() before
> schedule() it should be ok. Only the behaviour of code not holding
> kernel lock but global cli and calling schedule() has changed.

Agreed, however, there is one thing that has bugged me for a long time
(and which I believe is causing someone a problem at the moment) - when
we shut down a port, we're holding the BKL, and have global IRQs disabled.
We unhook the port from the serial drivers chain, and maybe free and
reclaim the IRQ with a different handler, and then disable the IRQ from
the port in question.

If we happen to schedule within request_irq, it doesn't take too much
imagination to see that Bad Things can happen.

(There is a report of complete lockup, and re-ordering stuff around here
fixes the problem, but the example patch changed a number of things, and
I'm trying to work towards a proper solution).

-- 
Russell King ([e-mail])                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html