[PATCH 5/5] arm/perfevents: implement perf event support for ARMv6

[linux@arm.linux.org.uk (Russell King - ARM Linux)]

On Tue, Jan 05, 2010 at 10:26:58PM +0000, Jamie Iles wrote:
> Hi Will,
> 
> On Tue, Jan 05, 2010 at 06:07:44PM -0000, Will Deacon wrote:
> > I've been trying to test your patches with a quad-core ARM 11MPCore on a
> > Realview PB11MP board.
> > 
> > Unfortunately, I occasionally experience a complete system hang during some
> > profiling runs. I don't think it's your fault however, as it can occur even
> > when monitoring only software events. I've managed to reproduce this on the 
> > tip/master branch and got the following information [I enabled lock debugging]:
> Could it be to do with the fact that perf_event_task_sched_in() expects
> interrupts to be disabled but on ARM we have __ARCH_WANT_INTERRUPTS_ON_CTXSW
> defined and therefore run with interrupts enabled? If so, I'm not sure what
> the fix is!
> 
> At the moment, ARM is the only platform that context switches with interrupts
> enabled and has perf event support.

If perf event support is only safe with interrupts disabled, it should
disable them.  Maybe a patch to do that conditional on
__ARCH_WANT_INTERRUPTS_ON_CTXSW would be more acceptable than an
unconditional one - don't know.

We could only define __ARCH_WANT_INTERRUPTS_ON_CTXSW for VIVT supporting
kernels, which is the reason for it existing (the interrupt latency for
VIVT would otherwise be unacceptable.)  This approach would mean that
perf events wouldn't be usable on VIVT CPUs (which includes Xscale CPUs.)