On 11/30/2018 11:19 PM, Rob Herring wrote:

On Fri, Nov 30, 2018 at 3:08 PM Pavel Machek [off-list ref] wrote:

quoted

Hi!

quoted

quoted

Pavel gave following examples:

eth0:green:link
adsl0:green:link
adsl0:red:error

So we would have e.g.:

associated-vl42-device = <&camera1>;
associated-network-device = <&phy1>;
associated-block-device = <&phy1>;

Variable property names are kind of a pain to parse.

Ok, would it be enough to have associated-device = <&whatever>?

Yeah, but I though you needed the device type name in there.

quoted

quoted

Perhaps when LEDs are associated with a device, we shouldn't care
within the context of the LED subsystem what the name is. The
association is more important and if you have that exposed, then you
don't really need to care what the name is. You still have to deal
with a device with more than 1 LED, but that becomes a problem local
to that device.

What I'm getting at is following a more standard binding pattern of
providers and consumers like we have for gpios, clocks, etc. So we'd
have something like this:

ethernet {
  ...
  leds = <&green_led>, <&red_led>;
  led-names = "link", "err";
};

Basically every single device could have a LED associated with it
("activity"). Would doing it like this mean we'd have to modify every
single driver to parse leds / led-names properties?

Normally, that's how properties like this would work. A driver is also
what knows how the leds should function. 

This is not true in case of associations where LED controller is
an independent device, as in Pavel's example [0].

A driver can retrieve the led
and associate it with the 'foo-bar' function. The 'foo-bar' function
then doesn't have to be defined in DT nor exposed to userspace. It
wouldn't even have to be driver specific. The driver's subsystem could
handle it all if the led functions are standardized. Though then you'd
be back to needing standard names for 'led-names', but that's no worse
that trigger names. This model would also allow getting rid of
'linux,default-trigger' properties in a lot of cases which wouldn't be
a bad thing.

However, having drivers handle this is not required. You can iterate
thru the tree for nodes with 'leds' and find the node which has a
phandle to the led node you care about. 

This way of discovering associations between devices and LEDs
would be more reliable than by devicename part of LED class device
as discussed previously.

However, I've just tried to verify how it works, but
I can't find the way to get the of_node phandle from sysfs.

The "of_node" link in per-device dirs in /sys/devices point
to the directories containing DT properties of the node, but
I see no way to obtain the node phandle.

Sure, it's not that efficient,
but it does work and it's only done once. Basically, as long as the
linkage is there, we can make it work. I think using
'associated-device' might work better for the current implementation
of Linux LED support, but leds/led-names would be more inline with
other DT bindings. The current Linux implementation shouldn't dictate
the binding design.

[0] https://lkml.org/lkml/2018/11/13/103

-- 
Best regards,
Jacek Anaszewski