Comment 91 for bug 1685442

> As I suspected before the problem is in the atomic conversion:
>
> commit 74c090b1bdc57b1c9f1361908cca5a3d8a80fb08
> Author: Maarten Lankhorst <email address hidden>
> Date: Mon Jul 13 16:30:30 2015 +0200
>

Okay, it seems the "vblank wait timed out on crtc 0" problem has
it's potential cause in a number of patches.

For your bisected patch I get

$git describe --contains 74c090b1bdc57
v4.3-rc1~20^2~40^2~11

which would mean that mainline Kernel 4.3.0 is the first affected one.

For my bisected patch I get

$git describe --contains edde361711ef8
v4.6-rc1~12^2~22^2~30

and indeed for me the problem first occurs in Linux 4.6.0.

Then we also have the report from https://bugs.freedesktop.org/show_bug.cgi?id=93782#c30 which says "the bug was not there with kernels 4.7.x, it came with 4.8.x."

It therefore seems that GM965 broke independently a number of times during the last year. As far as I understand, in all cases a failing load detection of intel_tv_detect() for svideo was responsible, although the details of that failure might be different for different people.

Given this situation, it seems that currently nobody really understands, how to do load detection for svideo on this old hardware anymore.

What is annoying is that the failing load detection also breaks hardware which does not even have a physical svideo port. Is there any evidence that svideo works at all with recent kernels? If not, it might be best to disable it by default. Alternatively, there could be a "i915_legacy" driver, which is kept frozen at the status of v4.2.x for those who depend on svideo.