[Intel-gfx] ACPI/i915: Cannot configure display brightness on Dell Latitude E6440

Pali Rohár pali.rohar at gmail.com
Tue Sep 23 22:06:02 CEST 2014


Hello,

after big changes in acpi video/i915 code I cannot change display 
brightness on my Dell Latitude E6440 with kernel 3.17-rc6. With 
kernel 3.13 everything worked fine.

More information about this problem:

For configuring brightness on Dell laptops there are 4 ways:
1) via acpi video driver
2) via dell-laptop driver
3) via i915 drm driver
4) from userspace with special dell SMI call
   (e.g with program dellLcdBrightness from libsmbios package)

Methods 2) and 4) are same, both making special SMI call and Bios 
handing this request (just 2 is from kernel and 4 from userspace)

Method 1) via acpi video driver working, but is not perfect. 
Driver can be used to change brightness (but only some levels, 
probably this depends on acpi/DSDT tables), but cannot be used to 
retrieve current brightness (when BIOS/SMI change brightness acpi 
driver report old incorrect value). So I prefer dell-laptop 
driver instead acpi video.

Method 3) working even with 3.17-rc6 kernel but because that 
backlight device exported by i915 is marked as raw, desktop 
programs prefer to use other devices.

Moreover it looks like that methods 1) 2) and 4) just forward 
request to method 3). So in any cased brightness is changed by 
i915 drm driver.

I'm not sure (correct me if I'm wrong!) but I think that intel 
i915 drm driver accept changes (file intel_opregion.c) only if 
acpi function acpi_video_verify_backlight_support() returns true.

Function acpi_video_verify_backlight_support() returns true iff:
function acpi_video_backlight_support() returns true AND at least 
one of these functions returns false:
acpi_osi_is_win8()
acpi_video_use_native_backlight() 
backlight_device_registered(BACKLIGHT_RAW)

On my notebook acpi_osi_is_win8() returns true (as is win8 
compliant), backlight_device_registered(BACKLIGHT_RAW) returns 
true as I'm using intel i915 drm driver with raw backlight device 
and acpi_video_use_native_backlight() returns true/false 
depending on "video.use_native_backlight" kernel param. Default 
is true.

So if I want to have working acpi video driver with display 
brightness support I need to boot kernel with param: 
"video.use_native_backlight=0". I tested it with kernel 3.17-rc6 
and this param really enabled display brightness support via acpi 
video driver -- which is good.

Driver dell-laptop creating backligh device for brightness 
control only if acpi_video_backlight_support() returns false. 
There is complicated condition for it and when kernel is booted 
with "video.use_native_backlight=0" that function returns true.

So conclusion is: With current code in kernel 3.17-rc6 it is not 
possible to control brightness of display with native driver 
dell-laptop on Dell Latitude E6440 (and probably on others 
too)!!!

And Because laptop is win8 compliant and you create decision to 
use native driver (instead acpi one) it is not possible to 
control display brightness without tweeks in kernel cmdline.

As I wrote I would rather to use native dell-laptop driver for 
controlling brightness, but it is not possible.

So how to solve this problem?

Quick solution would be to set use_native_backlight false for 
some Dell laptops which means, that acpi video will be used and 
in this case intel i915 driver will *not* drop backlight change 
request.

Another solution could be to disable check in dell_laptop driver 
and add use_native_backlight=0 to hooks. But this create two 
backlight interfaces (which is not good), but only way (for now) 
how to make dell_laptop working again.

Better and maybe only one proper solution would be to teach intel 
drm i915 driver to not drop backlight change request for Dell 
laptops (or all??). (This allows to work both acpi video and 
dell_laptop drivers without any change and with *any* value in 
param use_native_backlight). I think that problematic code is in 
function asle_set_backlight() in file intel_opregion.c (but I'm 
not sure). My idea is that "drop" event function it is caused by 
this commit 0b9f7d93ca6109048a4eb06332b666b6e29df4fe (but I'm not 
sure).

At least something must be done as I think that I'm not only one 
who has Dell laptop and brightness configuration is really 
important...

Thanks.

-- 
Pali Rohár
pali.rohar at gmail.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20140923/0a6d3d30/attachment.sig>


More information about the Intel-gfx mailing list