Re: [PATCH] Revert "drm/i915: fix vblank wait test condition"

[Sitsofe Wheeler <sitsofe@yahoo.com>]

On Sun, Sep 05, 2010 at 10:26:25AM +0200, Fabio Comolli wrote:
> 
> Well, this is very strange. I also own an EeePC 900 (i915 of course)
> and I don't have this problem at all, nor I need the patch. Actually I
> run a plain 2.6.36-rc3 and it works perfectly.
> 
> I use the latest userspace available, with xorg 1.9 and Intel video
> driver v. 2.12 (Archlinux testing repositories).
> 
> Which kernel / userspace are you using?

OK that's interesting - are you seeing correct xrandr output after a
suspend and resume too? I previously said that everything was working
perfectly a few weeks ago (only correct inputs were reported, only one
FIFO message in dmesg, no tearing on video, no cursor disappearing,
screen size was correct when X started, running xrandr doesn't make the
screen flash white etc) and added my tested-by line only to find out
that there were cases where phantom inputs were still being reported (
http://marc.info/?l=linux-kernel&m=128272404206696&w=2 ).

I'm using a 2.6.36-rc3 kernel (with the patch). I think userspace X is
less relevant as I can reproduce the false VGA detection without X
starting please see http://bugs.freedesktop.org/show_bug.cgi?id=29857
for more details).  I've tested the 2.12 intel drivers which you can add
via xorg-edgers (which I think was 1.8 but I'd have to check) and the
2.9 drivers which ship with Ubuntu 10.04 (which is xorg 1.7.6). The
newer X userspace pieces show a nice speedup in things like desktop
effects and textured video but if X doesn't run and inputs are being
mis-detected then I don't think it is the root of the problem...

I'm interested to know if it is just me seeing this issue. Did you even
have a problem with 2.6.36-rc2? Do you have any debugging options (e.g.
lock dep checking) on that would slow your kernel down and stop the
issue from ocurring?

-- 
Sitsofe | http://sucs.org/~sits/