Re: [REGRESSION/PATCH] acpi: blacklist win8 OSI for ASUS Zenbok Prime UX31A

[Felipe Contreras]

On Tue, Jul 30, 2013 at 6:13 PM, Rafael J. Wysocki <rjw@xxxxxxx> wrote:

>> If 0 turns the screen off with the intel driver, 0 should turn the
>> screen off with the ACPI driver, having inconsistent behavior
>> depending on which driver is used is a bug.
>
> The ACPI driver simply exposes and interface to interact with the AML methods
> in the BIOS directly.

No, the ACPI driver is exposing a backlight interface, which has a
defined stable API.

Documentation/ABI/stable/sysfs-class-backlight

Yes, the interface doesn't define what should happen at 0, that is a
bug in the interface definition.

*How* it achieves that is an implementation detail.

> Yes, this is a mistake and shouldn't be designed this way.
>
> However, incidentally, this makes backlight control work on your machine.
>
> Anyway, we need all backlight drivers to work consistently and don't tempt me
> to rip the ACPI driver entirely from the kernel for what it's worth.

Yes, they should work consistently, and go ahead, rip the ACPI driver,
*then* you'll see many more people complaining about the Linux kernel
breaking user-space, which should never happen. Mistakes happen, but
if you do this willingly and knowingly, I think there would be
repercussions for you.

> Yes, that will break backlight on your system and *then* you can complain to
> Linus if you wish.

It is already broken in v3.11-rc3, in fact I just booted that to try
it out and it booted with the screen completely black (fortunately I
knew exactly what to type to change that).

Apparently this commit also needs to be reverted: efaa14c (ACPI /
video: no automatic brightness changes by win8-compatible firmware).
In this machine it makes the backlight work again (without
acpi_osi="!Windows 2012"), but by doing so the ACPI driver also turns
off the screen completely at level 0. Also, each time I change the
backlight level from X, the screen blinks as if going 100%, 0%, and
then the desired level.

For this particular machine simply applying the attached patch would
solve all those regressions, but who knows in other machines, I think
it's safer to revert efaa14c.

-- 
Felipe Contreras
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/