Re: [PATCH] video: mxsfb: get supply regulator optionally

From
Stefan Agner <stefan@agner.ch>
Date
2016-11-09 00:22:58
Also in
linux-devicetree, lkml
Reply to
Tomi Valkeinen, 2016-09-07
Hi Tomi,

I vote to merge this patch, see below:

On 2016-09-07 00:20, Tomi Valkeinen wrote:
On 06/09/16 21:23, Stefan Agner wrote:
quoted
On 2016-09-06 01:21, Tomi Valkeinen wrote:
quoted
Hi,

On 04/09/16 07:26, Stefan Agner wrote:
quoted
The lcd-supply is meant to be optional, there are several device-
trees not specifying it and the code handles error values silently.
Therefor, avoid creating a dummy regulator (and the associated
warning) by using devm_regulator_get_optional.

While at it, document that fact also in the device-tree bindings.
The binding change looks correct, but using
devm_regulator_get_optional() does not sound correct.

devm_regulator_get_optional() is to be used when the device in question
truly can function without the power supply. But if the supply is there,
it's just not controlled by the SW, devm_regulator_get() is to be used.
The framebuffer device can even function without a display, no problem
there.. Probably not really useful...
Yes. Of course, the question then becomes, why is the fb driver even
dealing with the LCD's regulator. But yes, I know the answer: because
that's how it has been done =).
Agreed, this property really shouldn't be part of the display controller
node.

The device tree should describe the hardware, and that is clearly not
how the hardware is wired up...

The DRM solution to this is having a separate panel node with a
(mandatory) power-supply property.
quoted
devm_regulator_get creates a dummy regulator and a warning. Afaik, the
dummy regulator was meant to be as an aid during development, but not as
a permanent solution. This is what the initial commit of the dummy
regulator says:
Yep, the fixed regulator is afaik the correct solution to represent
non-controllable regulators.
quoted
quoted
In order to ease transitions with drivers are boards start using regulators
provide an option to cause all regulator_get() calls to succeed, with a
dummy always on regulator being supplied where one has not been configured.
A warning is printed whenever the dummy regulator is used to aid system
development.
I think we should either make the property mandatory and fix the device
trees or we should fix the driver to support an optional regulator. The
code already supports the reg_lcd being NULL, which is probably mostly
pointless right now as devm_regulator_get always returns a dummy
regulator.
To really clean this up, the LCD driver should be separated from the fb
driver. But that's pointless work on a framework that should be
deprecated (is there a DRM driver for this in the works? =).
Not that I am aware of, but I am considering it actually...
I'm fine with the _optional version, that's the easiest cleanup here.
And, I guess, it could be even argued that it's correct in some cases,
as the fb output could go outside the board, to some externally powered
display.

I'm fine with doing more cleanups too, if it eases the maintenance
burden in the future. But I don't see what the cleanups for the device
trees would really give us here.

Mark, what do you say?
So this is what mark said:

On 2016-09-12 16:47, Mark Brown wrote:
On Wed, Sep 07, 2016 at 10:20:25AM +0300, Tomi Valkeinen wrote:
quoted
I'm fine with the _optional version, that's the easiest cleanup here.
And, I guess, it could be even argued that it's correct in some cases,
as the fb output could go outside the board, to some externally powered
display.
I'd *prefer* to see the supplies being specified, if only for the
encouragement of the others.  But if you want to do the optional thing
anyway...
In this case we really don't want people encourage using this
property... Making it optional is not the solution, but a band aid until
we have a proper solution (and deprecate the property...)

--
Stefan
Thread (8 messages) 8 messages, 4 authors, last reply 2016-11-09
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help