On Fri, Nov 24, 2006 at 02:08:53AM +0100, Jesper Juhl wrote:

On 24/11/06, David Chinner [off-list ref] wrote:

quoted

On Thu, Nov 23, 2006 at 07:37:00PM +0100, Arjan van de Ven wrote:

quoted

On Thu, 2006-11-23 at 14:16 +0100, Ingo Oeser wrote:

quoted

Hi there,

David Chinner schrieb:

quoted

If the softirqs were run on a different stack, then a lot of these

softirqs DO run on their own stack!

So they run on a separate stack for 4k stacks on x86?

Yes, with 4K stacks there's sepperate IRQ stack.

Ok, thanks.

quoted

They don't run on a separate stack for 8k stacks on x86 -
Jesper's traces show that - so this may indicate an issue
with the methodology used to generate the stack overflow
traces inteh first place. i.e. if 4k stacks use a separate
stack, then most of the reported overflows are spurious
and would not normally occur on 4k stack systems..

Well, some of the traces show that we were down to ~3K stack free with
8K stacks, so ~5K used. Even with 4K stacks and sepperate stack for
IRQs we will still be uncomfortably close to the edge in those cases.

Sure - i didn't say there wasn't a problem - more just indicating
that most of the traces would not have happened on a 4k stack box so
it's harder to tell which of the traces you posted would actually
lead to an overflow.

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group