On 01/24/2013 04:30 PM, Jon Hunter wrote:

Hi Vaibhav,

On 12/31/2012 07:07 AM, Vaibhav Bedia wrote:

quoted

AM33XX has two timers (DTIMER0/1) in the WKUP domain.
On GP devices the source of DMTIMER0 is fixed to an
inaccurate internal 32k RC oscillator and this makes
the DMTIMER0 practically either as a clocksource or
as clockevent.

Currently the timer instance in WKUP domain is used
as the clockevent and the timer in non-WKUP domain
as the clocksource. DMTIMER1 in WKUP domain can keep
running in suspend from a 32K clock fed from external
OSC and can serve as the persistent clock for the kernel.
To enable this, interchange the timers used as clocksource
and clockevent for AM33XX.

I have been thinking about this some more. In the case where we are
using gptimers for both clock-events and clock-source (on both AM33xx
and OMAP) and I am wondering if it makes sense to switch the timers so
that we use the always-on timer for clock-source and a different one
from clock-events.

For OMAP, if we are not using the 32k-sync for clock-source, then we are
never going to achieve low power states during idle as we will always
have one gptimer running. And in this case, to your point below, it
would be better to use the always-on for clock-source so that in suspend
we can at least hit low power states and maintain time.

I have posted a patch today [1] that I hope will address this issue for
you. Can you give that a try?

Cheers
Jon

[1] http://www.spinics.net/lists/arm-kernel/msg221265.html