[Intel-gfx] Legacy backlight control with KMS and BACKLIGHT property

[Greg KH]

On Fri, Aug 14, 2009 at 01:34:20PM -0700, Keith Packard wrote:
> On Fri, 2009-08-14 at 13:10 -0700, Greg KH wrote:
> 
> > Why?  /sys/class/backlight should expose 2 backlights if they are known
> > and present, right?  Or is the problem that we don't expose new screens
> > in sysfs for userspace to be able to work out which is which?
> 
> We don't know which backlight goes with which LVDS device...

That's a problem, isn't there some way to figure it out from within the
kernel?

> > If so, then that should be fixed as well :)
> 
> It would be nice, but I haven't seen a proposal that attempts to address
> this. Hiding the backlight behind a DRM interface would mean that future
> fixes could be folded in without requiring changes in user-mode code.

But then what about all of the existing drivers already using the
/sys/class/blacklight interface?

Doesn't acpi export this information on lots of machines as well
already?

The only reason I had to write a new driver for the Samsung hardware,
was because they were not properly exporting the ACPI tables needed for
backlight support.  If they had done that, then no one would even be
having this conversation, which leads me to believe that lots of people
already depend on the /sys/class/backlight interface today, and that
will have to be resolved somehow no matter what.

thanks,

greg k-h