Bug#485465: battle for the brightness
reassign 485465 linux-2.6
retitle 485465 thinkpad-acpi: races on brightness change on a R50e
found 2.6.14-1
tag 485465 upstream
forwarded 485465 ibm-acpi@hmh.eng.br
owner ibm-acpi@hmh.eng.br
thanks
Synopsis: recently, brightness started acting up on R50e. Something is
racing with the firmware. Userspace attempts to support brightness are new,
and might be part of the equation, as misguided attempts to do it by
enabling thinkpad brightness hot-keys without fully disabling the BIOS can
make the problem far easier to hit. The sysfs backlight interface seems to
be more prone to trigger the race. A possible in-firmware race caused by
thinkpad-acpi is the main suspect at this point. That behaviour is present
since ibm-acpi 0.12a, kernel 2.6.14, but probably nothing was tripping it
before.
On Thu, 26 Jun 2008, jidanni@jidanni.org wrote:
> Ah, this workaround finally worked:
> # rmmod thinkpad-acpi; modprobe thinkpad-acpi brightness_mode=2
>
> Fn End typed after this apparently last even thru reboots!
Brightness settings on thinkpads ARE supposed to last through reboot and
power off, as long as you keep the NVRAM up-to-date. Thinkpad-acpi does so
in brightness_mode=2 and brightness_mode=3. It should always come back in
whatever brightness it was when you shut it down.
Do I understand correctly, and now that you're using brightness_mode=2, you
see no weird brightness changes when you lauch xdm, and your brightness keys
work just fine?
The R50e (from the ACPI tables you sent me) has exactly the same ACPI code
for brightness as the T43 and other ThinkPads of the T4x/R5x era. It
natively does what brightness_mode=2 does. The driver default,
brightness_mode=3, is something that I got from the previous maintainer of
ibm-acpi. It looks like it is time to rethink that code.
So, we have some sort of race, here. And brightness_mode=2 works around it.
I will look more into it, but unless I manage to reproduce this mess in my
T43, it won't be easy to root out the REAL reason for the breakage.
Jidanni, since brightness_mode=2 is working just fine for you, please
continue using it. It *is* supposed to work well on your box. In fact, one
of the things I will be trying to find out now is what kind of thinkpad
*requires* brightness_mode=3, because it looks like brightness_mode=2 is a
far saner default, even if that means accessing the CMOS NVRAM all the time
(which is expensive, and used to require disabling interrupts).
I will claim this bug for the thinkpad-acpi brightness change race, since
most of the baggage in it deals with that and your issue with brightness
"memory" across reboots seems to be solved.
For now, I will reassign the bug to the kernel, and mark it upstream (since
I am handling it, anyway).
--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh
Reply to: