Andi Kleen wrote:
>
> On Fri, Nov 03, 2000 at 09:37:40AM -0800, Matt D. Robinson wrote:
> > BTW, Andi, did 2.4 change the scheduler now so that you don't have
> > to get tasklist_lock to avoid having jobs scheduled underneath you
> > (in a panic()/interrupt state)?
>
> The current CPU cannot be rescheduled because panic is hanging in kernel
> mode. The panic also sends a stop IPI to the other CPUs, but until the
> IPI is processed there may be some scheduling. The IPI send function
> smp_call_function was also fixed to never schedule (it previously took
> a semaphore which sometimes could lead to the panic thread calling schedule)
> So it should be ok now.
>
> -Andi
Cool. We figured it was broken behavior -- we'd get messed up stack
pages for some dumps where scheduling took place. It was my
original understanding that this wouldn't happen, but then, 2.2 has
a number of broken issues.
--Matt
|