Re: [PATCH][GIT PULL] tracing: Fix compile issue for trace_sched_wakeup.c

[Ingo Molnar <mingo@elte.hu>]

* Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> > i.e. the 'fn(timer)' call crashed.
> 
> Right, and its doing an indirect function call from the first stack entry.. which 
> would seem to suggest someone scribbled our stack..

The ESP value itself seems generally sane:

> > [   11.980010] ESI: f65bdf50 EDI: f6806a00 EBP: f6806a30 ESP: c13dff04
> > [   11.980010] Process swapper (pid: 0, ti=c13de000 task=c13e2f20 task.ti=c13de000

As task.ti is 8K aligned (as it should be), and ESP is just a tad off the top of the 
stack (as it should be: a light timer irq interrupted the idle task).

The relevant stack section looks like this:

[   11.980010]  c103d297 00000000 c10460c1 c13dff4c ca105369 00000002 ffffffff 7fffffff
[   11.980010]  c103d52b ca105369 00000002 ca105369 0000002c f6806a00 00000000 f6806a04
[   11.980010]  ca105369 00000002 ca105369 00000002 00000000 f6805dac 00000000 c1420788

and that 00000000 value after c103d297 seems to be the called fn().

The stack content looks sane as well - we have the NULL entry, then various other, 
good looking entries that dont look corrupted per se. So either there's a 4 byte 
corruption of the stack, or the place where we took 'fn' from got corrupted: either 
the timer structure or the hrtimer per-cpu data area.

(More exotic crash scenarios as possible as well, if we consider .text corruption.)

Thanks,

	Ingo