Re: [PATCH v2] Input: silead - Add support for setting resolution based on dmi data

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Sun, Jan 22, 2017 at 12:03:07PM +0100, Hans de Goede wrote:
> Hi,
> 
> On 21-01-17 20:04, Dmitry Torokhov wrote:
> >Hi Hans,
> >
> >On Mon, Jan 02, 2017 at 11:55:38AM +0100, Hans de Goede wrote:
> >>On ACPI based tablets, the ACPI touchscreen node only contains info on
> >>the gpio and the irq, and is missing any info on the axis. This info is
> >>expected to be built into the tablet model specific version of the driver
> >>shipped with the os-image for the device.
> >>
> >>Add support for getting the missing info from a table built into the
> >>driver, using dmi data to identify which entry of the table to use and
> >>add info for the CUBE iwork8 Air and Jumper EZpad mini3 tablets on which
> >>this code was tested / developed.
> >>
> >>BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=187531
> >>Signed-off-by: Hans de Goede <hdegoede@redhat.com>
> >>---
> >>Changes in v2:
> >>-Put the dmi code in a separate silead_dmi.c file
> >>-Use device_add_properties to add the info
> >
> >This leaks data (device_add_properties() does deep copy of properties)
> >you unbind and rebind the driver.
> >
> >Also I would prefer this all be kept in the swamp that is platform
> >drivers and I believe we have enough infrastructure to do just that.
> >Could you please tell me if the following version of the patch works on
> >your devices?
> 
> I had to drop the unregistering of the bus-notifier, as the i2c controller
> driver may be a module (and in my case is as I'm testing with a Fedora
> distro kernel based .config) so the device may not get registered
> till after late_init.

Ah, indeed. I was bouncing between ACPI bus notifier and i2c one and
with i2c we indeed need the notifier to stick around.

Thanks.

-- 
Dmitry