On Wed, Aug 22, 2018 at 07:22:20AM -0700, Tony Lindgren wrote:

* J, KEERTHY [off-list ref] [180822 11:11]:

quoted

On 8/22/2018 2:13 PM, Johan Hovold wrote:

quoted

Yes, and a blacklist would make much more sense for something like this
if where talking about specific boards.

Black list is easier here?

After thinking about this a bit more I think the boards supporting
deep sleep should add a PM related dts property to enable deep sleep.

The board maintainers need to test and verify deep sleep for each
board, it's not something that just works for the SoC in general.
Some boards may use different powering for things like DDR where
it's power might be controlled by a GPIO regulator. And in some
cases deeper idle states may depend also on the PMIC being used.

Maybe we already have some dts property we can use to describe
the idle states the board hardware supports?

Yeah, unless you can infer this from an existing tree I guess you need
to add a new property. And indeed, a driver blacklist would suffer from the
same fundamental problem (with an ever expanding list of machines) as a
whitelist even if it would avoid regressing currently working systems.

Johan