Re: [PATCH 00/10] platform/x86: toshiba_acpi: move acpi add/remove to device-managed routines

[Alexandru Ardelean]

On Mon, 29 Mar 2021 at 15:38, Jonathan Cameron <jic23@xxxxxxxxxx> wrote:
>
> On Wed, 24 Mar 2021 14:55:38 +0200
> Alexandru Ardelean <aardelean@xxxxxxxxxxx> wrote:
>
> > This changeset tries to do a conversion of the toshiba_acpi driver to use
> > only device-managed routines. The driver registers as a singleton, so no
> > more than one device can be registered at a time.
> >
> > My main intent here is to try to convert the iio_device_alloc() and
> > iio_device_register() to their devm_ variants.
> >
> > Usually, when converting a registration call to device-managed variant, the
> > init order must be preserved. And the deregistration order must be a mirror
> > of the registration (in reverse order).
> >
> > This change tries to do that, by using devm_ variants where available and
> > devm_add_action_or_reset() where this isn't possible.
> > Some deregistration ordering is changed, because it wasn't exactly
> > mirroring (in reverse) the init order.
> >
> > For the IIO subsystem, the toshiba_acpi driver is the only user of
> > iio_device_alloc(). If this changeset is accepted (after discussion), I
> > will propose to remove the iio_device_alloc() function.
> >
> > While I admit this may look like an overzealous effort to use devm_
> > everywhere (in IIO at least), for me it's a fun/interesting excercise.
> hmm. I am dubious about 'removing' the support for non devm_ in the long
> run because it can lead to requiring fiddly changes in existing drivers
> (like this one :) and I don't want to put that barrier in front of anyone
> using IIO.

Yeah.
I also feel that the current driver is a bit fiddly.
I was undecided [when doing the series], whether to send it as a
whole, or start with sending just a few patches that make sense on
their own.

I might go via the second route and send these individually.

>
> However, I'm more than happy to see them used in very few drivers and
> nice warning text added to suggest people might want to look at whether
> then can move to a device managed probe flow
>
> Jonathan
>
> >
> > Alexandru Ardelean (10):
> >   platform/x86: toshiba_acpi: bind life-time of toshiba_acpi_dev to
> >     parent
> >   platform/x86: toshiba_acpi: use devm_add_action_or_reset() for
> >     singleton clear
> >   platform/x86: toshiba_acpi: bind registration of miscdev object to
> >     parent
> >   platform/x86: toshiba_acpi: use device-managed functions for input
> >     device
> >   platform/x86: toshiba_acpi: register backlight with device-managed
> >     variant
> >   platform/x86: toshiba_acpi: use devm_led_classdev_register() for LEDs
> >   platform/x86: toshiba_acpi: use device-managed functions for
> >     accelerometer
> >   platform/x86: toshiba_acpi: use device-managed for wwan_rfkill
> >     management
> >   platform/x86: toshiba_acpi: use device-managed for sysfs removal
> >   platform/x86: toshiba_acpi: bind proc entries creation to parent
> >
> >  drivers/platform/x86/toshiba_acpi.c | 249 +++++++++++++++++-----------
> >  1 file changed, 150 insertions(+), 99 deletions(-)
> >
>