[Intel-gfx] [PATCH v2] pwm: lpss: Make builtin so that i915 can find the pwm_backlight

[Hans de Goede]

Hi,

On 20-01-17 11:42, Thierry Reding wrote:
> On Fri, Jan 20, 2017 at 11:18:29AM +0100, Hans de Goede wrote:
>> Hi,
>>
>> On 20-01-17 10:55, Andy Shevchenko wrote:
>>> On Fri, 2017-01-20 at 10:48 +0100, Hans de Goede wrote:
>>>> I'm fine with doing a v3 with a comment, how about putting that
>>>> comment
>>>> right at all the module* stuff and explain there that that is to
>>>> stay as the builtin only status is meant to be temporary ?
>>>
>>> Can we do other way around? I mean that either i915 selects PWM_LPSS to
>>> be built-in, or uses request_module() call?
>>
>> PWM_LPSS needs to be built-in if enabled, a stripped down kernel
>> for non cherrytrail hardware does not need it ...
>>
>> Also (and esp for request_module) this means building knowledge
>> into the i915 driver about which pwm hardware there is on which
>> boards which is undesirable.
>>
>> Thierry, this does give me an idea though, what if we extend
>> the info passed to pwm_add_table with a module-name and
>> make pwm_get call request_module() ?
>
> I'm not sure that's even necessary. request_module() forwards the string
> you pass to it to the userspace helper, so you can pass things like the
> modalias to it. I suspect that for ACPI the modalias could be trivially
> derived from the provider name already in the table.

As you can see in the patch-set I've just send I've chosen to go with the
module_name in pwm_lookup. You're right that simply doing:

	module_request("acpi:%s", chosen->provider);

Would work, but I don't like having the "acpi:%s" bit in the pwm-core,
esp. not since someone will them come along and add support for i2c.
pci, etc. So just adding a module_name field seems saner.

Regards,

Hans