Hi,

On Thursday 26 March 2015 04:19 AM, Dmitry Torokhov wrote:

On Thu, Mar 26, 2015 at 04:09:23AM +0530, Kishon Vijay Abraham I wrote:

quoted

Hi,

On Thursday 26 March 2015 03:47 AM, Dmitry Torokhov wrote:

quoted

On Wed, Mar 25, 2015 at 3:11 PM, Kishon Vijay Abraham I [off-list ref] wrote:

quoted

Hi,

On Saturday 21 March 2015 02:55 AM, Arun Ramamurthy wrote:

quoted

Multi-port phy's may have per-port power supplies. Let's change phy core
to first attempt to look up the supply at the port level, and then, if
not found, check parent device.

Why not just have every port provide the power supply if it needs?
I don't think checking for parent device should be present in the phy-core at
all.

We need to do that if we want to keep compatibility with the current
DTSes: before this patch the supply would be always looked up on
device and not port level.

ah okay.
so just using regulator_get_optional(&phy->dev, "phy"); should be sufficient

This is for regulators specified at port level (&phy->dev represents
port).

quoted

right? Why do we need regulator_get_optional(phy->dev.parent, "phy");?

This is for compatibility with old multi-port bindings where supply is
specified at parent device level and phy_create() is called with dev and
node that is not NULL and not the same as dev->of_node. I have no idea
if such bindings exist in wild, but wanted to keep them working given
stated DT stability rules.

Such a binding doesn't exist. So let's keep only the
regulator_get_optional(&phy->dev, "phy"); part. Only TI SoCs and recently
sun9i started using phy-supply and none of them use multi-phy PHY provider.

Thanks
Kishon