Messages in this thread | | | Date | Tue, 11 Mar 2008 14:58:30 +0100 | From | "Zdenek Kabelac" <> | Subject | Re: [PATCH REPOST] Run IST traps from user mode preemptive on process stack |
| |
2008/3/11, Andi Kleen <andi@firstfloor.org>: > > I'm not sure whether this is good or bad sign - but with this patch > > you have posted, > > I do not have to wait in qemu for minutes to get the 'busy-loop' - I > > get the exact same loop almost instantly when I start disk read of LV > > partition and running my simple module insertion testcase. > > > Hmm, my fix was intended to fix a gdb problem. Or rather the gdb > problem was already fixed by disabling some functionality and with > this patch the functionality would be reenabled again. > > If it fixed a qemu problem too that's great but was unintended > on my part. In fact it is a little worrying because there should > be no user visible change. Can you double check by unapply/test/reapply/test > the patch really makes a difference?
I've not said it has 'fixed' my problem - I said it has exposed the problem much faster. But now when I've made a double test - I think I was wrong - I've been probably judging too fast and its also probably depending on the sunlight position, but now I can see that even with recent git kernel without your patch I get the busy loop pretty quick :(.
> > > This patch changes these traps instead to always switch > > > to the process stack when the trap originated from user mode. > > > For kernel traps it keeps running non preemptive on the IST stack > > > because that is much safer (e.g. to still get nmi watchdog events > > > out even when the process stack is corrupted) > > > > > > > So what should I check to get fixed my problem in qemu ? > > > I don't know. Someone has to debug it.
But who :) ?
Zdenek
| |