On Mon, Oct 03, 2011 at 11:57:03AM +0100, Lorenzo Pieralisi wrote:

On Sat, Oct 01, 2011 at 04:06:46PM +0100, Russell King - ARM Linux wrote:

quoted

On Fri, Sep 30, 2011 at 02:42:53AM -0700, Barry Song wrote:

quoted

From: Barry Song <redacted>

we save the l2x0 registers at the first initialization, and platform codes
can get them to restore l2x0 status after wakeup.

Cc: Lorenzo Pieralisi <redacted>
Signed-off-by: Barry Song <redacted>
Reviewed-by: Santosh Shilimkar <redacted>
Tested-by: Shawn Guo <redacted>

This looks fine, thanks Barry.

The one remaining issue is whether we use this or use Lorenzo's patches.
I feel that Barry's version is a lot simpler and easier to use, so this
is my preference.  Anyone else got another opinion between the two?

Basically the two patches do the same thing except that Barry's generic
resume code cannot be called before the MMU is turned on. On the other hand
Barry's patch is certainly simpler to maintain and I think it should be 
merged, preference shared; I would not say it is easier to use though, usage 
is identical.

It is agreed, better to have it in C than in asm for maintenance reasons, 
asm can turn out overkill.
If there is a compelling case for generic L2 asm we can always add it at a 
later stage (at least Origen, latest iMX and I think Tegra will need early L2
resume from idle).

Ok, so it sounds like we're settled on Barry's code... so can it head to
the patch system please?