Buggy BIOSes and intel driver.

[Raúl Sánchez Siles]

Hello All:

If you have followed bug #11432
(https://bugs.freedesktop.org/show_bug.cgi?id=11432) you could conclude
that some BIOSes are so buggy that prevents intel driver do its job
properly causing some hard freezes.

As per discussion this happens when pipe A is disabled by the driver when
necessary in order to save some energy. Well, the first option that comes
to mind to avoid such freezes is not disabling the pipe. That's what I've
tried with the patch I've attached there, from what I welcome comments, of
course.

This is not the optimal solution, IMHO since the hack consist on telling
Xserver that a monitor is indeed detected on that pipe. This has a lot of
drawbacks. Maybe even doing the hack this way is not the most advisable
thing to do, but it was the only one I manage to get so far. After my
computer is not freezing like that anymore.

Some possible solutions are:
  1-Enable pipe A in a less intrusive way. Pretending we always have VGA
monitor attached is anything but smart. Your DM will always think you are
dealing with 2 screens which may turn it crazy. Forcind pipe A enabled or
not disabling it upon the ForceEnablePipeA option is preferred IMHO to use
the option to fake VGA detection. Working on this currently. I think intel
driver commit b31bef1a8effa9acb6de7edd206b9d8c48d88144 has anything to do
with this.
  2-Using xrandr to send an on command same the --off command is send. I
don't really know the implications of this. Comments appreciated.
  3-Use pipe A to output LVDS. Altough this is possible theoretically as per
Keith words, I've been unable to test this is indeed true. Have to do some
research yet.
  4-Treat BIOS in a way such the problem is avoided. It would be the kernel
that should provide this, so this is not in the xorg reach. All the needed
information should be gathered and send to the linux-acpi group.

  I think this is enough for a first post. But I have lots of doubts on most
fields. I'll hold on my horses till I have some feedback either from you of
from any test.

  As you may have noticed I guess this may not be an Xorg priority. Still I
think this is the ML that could help better into having a bunch of even
happier users ;)

  Thanks all.

-- 
     Raúl Sánchez Siles
----->Proud Debian user<-----
Linux registered user #416098