Re: [PATCH v1 09/11] thermal/core: Add a linked device parameter

[Rafael J. Wysocki]

On Tue, Apr 4, 2023 at 9:01 PM Daniel Lezcano <daniel.lezcano@xxxxxxxxxx> wrote:
>
> On 27/03/2023 18:16, Rafael J. Wysocki wrote:
> > On Tue, Mar 7, 2023 at 2:38 PM Daniel Lezcano <daniel.lezcano@xxxxxxxxxx> wrote:
> >>
> >> Some drivers want to create a link from the thermal zone to the device
> >> sysfs entry and vice versa.
> >
> > Which device is this, exactly?
>
> ls -alh /sys/bus/acpi/drivers/thermal/LNXTHERM:00/
>
> [ ... ]
>
> lrwxrwxrwx    1         0 thermal_zone ->
> ../../../virtual/thermal/thermal_zone0
>
> [ ... ]
>
> The ACPI driver is the only one doing this AFAICT.
>
>
> >> That is the case of the APCI driver.
> >>
> >> Having a backpointer from the device to the thermal zone sounds akward
> >> as we can have the same device instantiating multiple thermal zones so
> >> there will be a conflict while creating the second link with the same
> >> name. Moreover, the userspace has enough information to build the
> >> dependency from the thermal zone device link without having this cyclic
> >> link from the device to thermal zone.
> >>
> >> Anyway, everything in its time.
> >>
> >> This change allows to create a these cyclic links tz <-> device as
> >> ACPI does and will allow to remove the code in the ACPI driver.
> >
> > Well, I'd rather have it in the driver than in the core TBH.
> >
> > If ACPI is the only user of this, let it do the dirty thing by itself.
> >
> > There are two cases which would justify making this change:
> > 1. There will be more users of it going forward (seems unlikely from
> > the description).
> > 2. It gets in the way of some other changes somehow.
> >
> > I kind of expect 2. to be the case, so how does it get in the way?
>
> Shall we do the same approach than 'Menlow' and add an option to remove
> the thermal zone link in the acpi sysfs directory ?

That depends on the answer to the question above.