Re: Panic on dom0 shutdown

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Jan 09, 2013 at 09:56:30AM +0100, Johan Ihrén wrote:
> Hi,
> 
> > OK, so can you try the attached patch (in addition to building with
> > -fno-optimize-sibling-calls) ?
> 
> Smallest patch I ever saw, but I understand what you're trying to achieve 
> (forcing pirq_interrupt to not have a tail-call to whatever it is calling). 
> However, again, there is no change to the resulting trace at the point of the 
> panic:
> 
> > Stopped in pid 745.1 (halt) at netbsd:bus_space_read_4*0x8:
> > bus_space_read_4()
> > pirq_interrupt()
> > Xresume_xenev6() at netbsd:Xresume_xenev6+0x47
> 
> I don't know anything about how the trace function is implemented, but 
> obviously adding a dummy statement to the tail end of pirq_interrupt didn't 
> help us. So I added a dummy function between pirq_interrupt and the call to 
> the function pointer to see where the trace starts to present the correct 
> call chain.
> 
> This changed the trace to: 
> 
> > Stopped in pid ...
> > bus_space_read_4()
> > johani_dummy()
> > pirq_interrupt()
> > Xresume_xenev6() at netbsd:Xresume_xenev6+0x47
> 
> which tells me that from the POV of getting a correct trace our problem is 
> not with getting rid of a tail-call from pirq_interrupt, but rather getting 
> rid of a tail-call from the function that the function pointer refers to.

What does johani_dummy() do ? If it's just an empty function, I don't
understand how we can have such a call trace.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--