On 11/10/11 10:33, David Brown wrote:

On Tue, Nov 08, 2011 at 10:34:02AM -0800, Stephen Boyd wrote:

quoted

Currently the MSM timers use the same physical counter
for the clockevent and clocksource. This works as long
as the clocksource isn't stopped from ticking during normal
operation but unfortunately that isn't the case and the
clocksource is stopped when the clockevent is shutdown.
Even worse, switching the clocksource via sysfs at runtime will
hang the system.

This series reorganizes the MSM timer code so that one counter
is only used for either a clocksource or a clockevent, and not 
both. In the process we reduce the lines of code and fix
a few long-standing bugs.

How do you test these on the upstream kernel?  Booting successfully is
probably a good starting point, though.

Booting is good because that means you're getting timer interrupts.
After that, do some sleep 10, 20, 30, etc. tests with a stopwatch to
make sure that time isn't completely off.

You can also check bogoMIPS and see what the before and after is. Beyond
that I suppose some kind of ntp or gettimeofday test would be good.

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.