[Letux-kernel] regression: drm/omapdrm: Move commit_modeset_enables() before commit_planes()

[Tomi Valkeinen]

On 15/04/17 10:44, Andreas Kemnade wrote:
> Hi Jyri,
> 
> I am often seeing only some moving horizontal white strips (maybe 1-2
> pixels thick) when I boot my gta04 (omap 3 dm3730 + panel_tpo_td028ttec1
> + encoder_opa362).
> 
> After a git bisect session, the verdict is: this patch is guilthy:
> 
> commit 897145d0c7010b4e07fa9bc674b1dfb9a2c6fff9
> Author: Jyri Sarha <jsarha at ti.com>
> 
>     drm/omapdrm: Move commit_modeset_enables() before commit_planes()

I was able to reproduce on my very old omap3 beagleboard. Looks very
much like a omap3 DSS HW bug to me. The unfortunate thing is that even
reverting the patch won't remove the issue, if the userspace happens to
configure things in certain order.

It looks to me that if the DSS output is enabled without an overlay (the
primary plane, normally), DSS gets stuck into a bad state, and only
reboot helps (because we can't reset, there's no omap hwmod reset
framework).

But there should be nothing wrong with that setup, and if there's first
been a successful display enable, then later we can enable the DSS
without any planes.

So... At the moment, I have no idea how to fix it. Reverting the patch
will cause other things to start breaking, but perhaps that's the only
option as a short term kind-of-fix.

 Tomi

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://lists.goldelico.com/pipermail/letux-kernel/attachments/20170419/1f3a2c91/attachment.asc>