Re: [RFC PATCH 02/16] ASoC: pcm512x: use "sclk" string to retrieve clock

[Mark Brown <broonie@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 1244 bytes --]

On Tue, Apr 14, 2020 at 04:02:00PM -0500, Pierre-Louis Bossart wrote:

> Thinking a bit more on this, is the objection on the notion of using a fixed
> string, on the way it's registered or the lack of support for clocks in
> ACPI?

The issue is using a clock named in the global namespace.  Like I keep
saying you're not using ACPI here, you're using board files and board
files can do better.

> From a quick look, the use of a fixed string is rather prevalent, see below.
> Less than 10% of codec drivers rely on a NULL string, so is it really a
> dangerous precedent so use "sclk" in this case? It seems to me that all clk
> providers need to use a unique string - what am I missing here?

> adau17x1.c:	adau->mclk = devm_clk_get(dev, "mclk");

Notice how all the clock lookup functions take both a device and a
string - the device is important here, the string is namespaced with the
device in most usage (including board file usage) so if two different
devices ask for the same name they might get different clocks.

> wm8962.c:	pdata->mclk = devm_clk_get(&i2c->dev, NULL);

This is how lookups that don't even specify a name can work.  You seem
to want to rely on the name only which is very much not good practice,
even on board files.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]