Re: [PATCH] ACPI: battery: register power_supply subdevice even when battery not present

[Alan Jenkins <sourcejedi.lkml@googlemail.com>]

On 9/11/09, Anton Vorontsov <avorontsov@ru.mvista.com> wrote:
> On Fri, Sep 11, 2009 at 01:28:42PM +0100, Alan Jenkins wrote:
>> [CC power_supply maintainers]
>>
>> I tried to fix the ACPI battery driver so it created a power supply
>> device even if the battery is not present.  All the other battery
>> drivers do this.  It would let e.g. gnome-power-manager to detect the
>> presence of a battery bay, and allow configuration of battery
>> behaviour even when the battery has been removed.
>
> This was discussed several times already.
>
> http://lkml.org/lkml/2009/4/27/314
> http://lkml.org/lkml/2007/10/27/170
>
>> Unfortunately this is not as easy as I thought.
>
> And not possible for certain batteries.
>
> [...]
>> I can think of some more complex ways to do this
>>
>> 1) Destroy and recreate the battery device on hotplug.
>> 2) Modify the generic power supply class to allow changing the set of
>> attribute.
>> 3) Restructure the interface to provide a "battery bay" device as a
>> parent.
>>
>> 1) and 2) are hacks.  1) could at least cause annoyingly spurious UI
>> events.  2) sounds like a bad idea, but existing userspace _might_
>> handle it because it already happens at registration time.  (Power
>> supply attributes aren't available on the initial ADD uevent; they are
>> added in a follow-up CHANGE uevent).
>>
>> 3) could be backwards compatible and relatively straightforward.  The
>> "battery bay" could be a new type of power supply device, with no
>> attributes of its own.
>
> Yes, that would be nice. There could be some attributes though,
> I can think of a temperature sensor (measures ambient/bay
> temperature), a battery presence/lock sensor etc.
>
> IIRC, battery bay in iPaq hx4700 devices can report if a battery is
> locked (that's all it can do, but still useful).
>
>> It could be created automatically for non-acpi batteries.
>
> Why? If there is no bay, we don't need any device for it.

I thought it would be easier for userspace... but actually there's no
point since programs will still have to handle older kernels.

Ok, I'll write 3) and see what happens.

Alan