Re: stack dumps, CONFIG_FRAME_POINTER and i386 (was Re: sysrq shows impossible call stack)

[Andrew Morton <akpm@osdl.org>]

Adam Litke <agl@us.ibm.com> wrote:
>
> On Tue, 2004-04-20 at 18:41, Andrew Morton wrote:
> > Andrew Morton <akpm@osdl.org> wrote:
> > >
> > > Roland Dreier <roland@topspin.com> wrote:
> > >  >
> > >  >     Adam> This problem was annoying me a few months ago so I coded up
> > >  >     Adam> a stack trace patch that actually uses the frame pointer.
> > >  >     Adam> It is currently maintained in -mjb but I have pasted below.
> > >  >     Adam> Hope this helps.
> > >  > 
> > >  > Thanks, that looks really useful.  What is the chance of this moving
> > >  > from -mjb to mainline?
> > > 
> > >  Good, but it needs to be updated to do the right thing with 4k stacks when
> > >  called from interrupt context.
> 
> The show_trace() for the CONFIG_FRAME_POINTER case will now be called
> the same way as the existing code.

I still don't see any code in there to handle the transition from the
interrupt stack page to the non-interrupt stack page in the 4k-stacks case?

> This brings up a question though. 
> It doesn't appear to me that anyone is actually calling
> show_trace_task() yet.  Am I missing something or should we change all
> the callers of show_trace() to use show_trace_task()?

You're right - we've killed off all of its callers.  Neat.  I shall
administer the coup de grace.