Re: [PATCH 2.6.29] eeepc-laptop: report brightness control events via the input layer

[Alan Jenkins <alan-jenkins@tuffmail.co.uk>]

Corentin Chary wrote:
> On Mon, Jun 8, 2009 at 5:24 PM, Alan
> Jenkins<sourcejedi.lkml@googlemail.com> wrote:
>   
>> On 4/4/09, Matthew Garrett <mjg59@srcf.ucam.org> wrote:
>>     
>>> On Fri, Apr 03, 2009 at 06:57:50PM +0100, Darren Salt wrote:
>>>       
>>>> This maps the brightness control events to one of two keys, either
>>>> KEY_BRIGHTNESSDOWN or KEY_BRIGHTNESSUP, as needed.
>>>>
>>>> Some mapping has to be done due to the fact that the BIOS reports them as
>>>> <base value> + <current brightness index>; the selection is done according
>>>> to
>>>> the sign of the change in brightness (if this is 0, no keypress is
>>>> reported).
>>>>
>>>> (Ref.
>>>> http://lists.alioth.debian.org/pipermail/debian-eeepc-devel/2009-April/002001.html)
>>>>
>>>> Signed-off-by: Darren Salt <linux@youmustbejoking.demon.co.uk>
>>>>         
>>> The reason I didn't do this is that the Eee changes the input brightness
>>> in hardware, which means reporting it via the input layer as well can
>>> cause a single keypress to raise the brightness by two steps - one in
>>> hardware and one triggered by userland's response to the key press. I'd
>>> be a little bit wary of this causing problems.
>>>
>>> On the other hand, the default behaviour of the acpi video driver is to
>>> change the brightness itself and then also to send the even to
>>> userspace, so I guess if it was going to break things it probably would
>>> have done already...
>>>       
>> Actually, I think userspace has learnt to hack around it but it
>> doesn't work perfectly.  I would like to request that this change be
>> reverted, or otherwise improved.
>>
>> Before this patch (2.6.29.4), gnome-power-manager doesn't interfere
>> with the brightness keys, and they work smoothly.
>>
>> After this patch (2.6.30-rc7), g-p-m produces a "nice" popup in the
>> middle of my tiny netbook screen.  Unfortunately it can't be disabled,
>> but that's not your fault :-).  The brightness controls generally work
>> ok.  It doesn't jump two steps in response to one brightness keypress.
>>  But:
>>
>> 1) If I'm thrashing the SSD.  I get jerky after-effects, where g-p-m
>> seems to take too long to "catch up" with the brightness change.
>>     
>
> There is the same "lag" problem with sound :/
>   

Yeah :/.  At it's worst, it isn't a pure "lag".  It's a bit harder to 
explain.  The firmware still changes the brightness immediately.  It 
seems that when g-p-m gets delayed, it responds _wrongly_. It doesn't 
realize that the firmware already changed the brightness, so it changes 
the brightness again.

That's why I think this is a bad "feature".  User-space is working 
around it, but the workaround isn't reliable.  I think the proper 
solution is that if userspace wants to respond to firmware-initiated 
brightness changes, it should listen for uevents on the brightness class 
device.

You can see the problem most clearly by pressing the brightness keys in 
quick succession e.g. 3 times in a row, and see 3+3 brightness changes.  
I reproduced this with

1) 2 writers:
    F=1; while true do dd if=/dev/zero of=$F bs=1M count=1 conv=sync; done &
    F=2; while true do dd if=/dev/zero of=$F bs=1M count=1 conv=sync; done &

2) 1 reader:
    dd if=/dev/sda of=/dev/null

3) Drop caches before pressing the brightness keys
    echo 1 > /sys/proc/vm/drop_caches


>> 2) If I go all the way down from full (holding down the "brightness
>> down" key), and then back up a few steps.  I get a noticable flash
>> where the brightness looks to go up two steps, then down one.  It's
>> probably most noticable here because the step change between the
>> lowest and the second lowest brightness is much more visible than any
>> of the other steps.
>>
>>     
> I tried to install gnome-power-manager to test that, but there is no "popup".
> What do I have to install to test that ? entire gnome desktop :/ ?
>
> Thanks
>   

That's weird.  I'm running it from KDE here (g-p-m package version 
2.22.1-4 on debian stable).  I'm pretty sure the only other GNOME app I 
have installed is Pidgin.  I would know if I had much more installed, 
because I'm often running short of disk space :-).

Maybe you have a newer version, that doesn't try to do such unreliable 
things?

Thanks for your time
Alan