Mark,

     Thanks for your reply and advice.

On 08/28/2014 11:11 PM, Mark Rutland wrote:

On Thu, Aug 28, 2014 at 10:17:58AM +0100, Mark Rutland wrote:

quoted

Hi Kever,

On Thu, Aug 28, 2014 at 02:40:17AM +0100, Kever Yang wrote:

quoted

We need use the hrtimer, which need the arch-timer to be 'always-on'

I asked a question on the last posting [1]. Can you please confirm
either way?

Thanks,
Mark.

[1] lists.infradead.org/pipermail/linux-arm-kernel/2014-August/282327.html

To clarify: if there are low power states that the CPU can enter where
we lose state, then this patch isn't correct.

rk3288 has low power state and may turn off the cpu power domain
which will lost any logic state in cpu.

A more general approach would be to enable the broadcast hrtimer for
arm, as has been done for arm64.

See commit 5d1638acb9f6 (tick: Introduce hrtimer based broadcast) which
introduced the broadcast hrtimer, and commit 9358d755bd5c (arm64:
kernel: initialize broadcast hrtimer based clock event device) which
added the requisite plumbing for arm64.

I'll going to implement this and send another patch.

Thanks.

-Kever