Re: [BUG] kprobes crashing because of preempt count

[Ananth N Mavinakayanahalli <ananth@in.ibm.com>]

On Fri, Jul 01, 2011 at 09:01:00PM +0900, Masami Hiramatsu wrote:
> (2011/07/01 20:36), Ananth N Mavinakayanahalli wrote:
> > On Fri, Jul 01, 2011 at 10:12:03AM +0900, Masami Hiramatsu wrote:
> >> (2011/06/30 22:23), Steven Rostedt wrote:
> > 
> > ...
> > 
> >>> Do we really need to have preemption disabled throughout this? Is it
> >>> because we don't want to migrate or call schedule? Not sure what the
> >>> best way to fix this is. Perhaps we add a kprobe_preempt_disable() that
> >>> is checked as well?
> >>
> >> I think the best way to do that is just removing preemption disabling
> >> code, because
> >> - breakpoint exception itself disables interrupt (at least on x86)
> >> - While single stepping, interrupts also be disabled.
> > 
> > On 64-bit powerpc, kprobe handlers are run with interrupts enabled
> > (MSR_EE = 1), but most instructions (including loads/stores) are
> > emulated, so for the most part, we don't take the sstep exception.
> 
> Yeah, it seems that same thing is done on arm too. And I'm sure that
> However, I'm still not sure that entire int3 exec path can run without
> calling inc_preempt_count.
> It seems that the function is very primitive, and I doubt we can
> allow to put kprobes on that...

Right. I think all preempt manipulation routines need to be __kprobes.

Also, Steve is testing the -rt tree where artefacts related to
locking/preemption are possibly quite different from the mainline that
may also be at play here.

Ananth