Bug 71431

Summary: radeon graphics stopped working since kernel 3.10 on AMD a4-3300
Product: DRI Reporter: ka.nick
Component: DRM/RadeonAssignee: Default DRI bug account <dri-devel>
Status: RESOLVED MOVED QA Contact:
Severity: critical    
Priority: medium CC: ka.nick
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg output
none
Xorg.0.log
none
lspci output
none
bisect log with some comments added
none
possible fix none

Description ka.nick 2013-11-09 16:08:17 UTC

    
Comment 1 ka.nick 2013-11-09 16:27:00 UTC
Created attachment 88936 [details]
dmesg output

X11 doesn't work on my dual-graphics AMD laptop with kernels 3.10.X - 3.12.X, on 3.9.X and earlier kernels does.
Upon startup lxdm shows a garbage instead of normal picture (but the colors are from the picture), X restarts every few seconds, but kernel itself works fine - I can ssh into the comp or restart using the magic key.
Have been trying different combinations: EXA/Glamor, various xorg options with basically the same result: in some cases I can see the normal working mouse over the garbage, in some cases I get "GPU stall" message in dmesg, or EQ overflow in Xorg.0.log...
R600_DEBUG=sdsafemath - improves a picture a bit;
ColorTiling off - makes things even better - in this case about 70% of the screen are drawn correctly, but at wrong position on the screen;
NoAccel on - works as with 3.9.X kernel - windows and lxdm screen - ok, background with errors.
Comment 2 ka.nick 2013-11-09 16:28:11 UTC
Created attachment 88937 [details]
Xorg.0.log
Comment 3 ka.nick 2013-11-09 16:31:19 UTC
Created attachment 88939 [details]
lspci output
Comment 4 Alex Deucher 2013-11-10 13:41:07 UTC
(In reply to comment #1)
> X11 doesn't work on my dual-graphics AMD laptop with kernels 3.10.X -
> 3.12.X, on 3.9.X and earlier kernels does.

Can you bisect with git?
Comment 5 ka.nick 2013-11-11 08:20:37 UTC
Yes, within few  days...
Comment 6 ka.nick 2013-11-14 07:57:50 UTC
Created attachment 89177 [details]
bisect log with some comments added

Alex,
here is a bisecting results. 
Please note that few commits away from the first bad commit it became half-broken - background isn't corrupted but lxdm doesn't start, and x keeps restarting.
Comment 7 ka.nick 2013-11-21 08:13:54 UTC
Alex?
Did I provide all necessary information, don't I?
Comment 8 Alex Deucher 2013-11-21 16:25:04 UTC
(In reply to comment #7)
> Alex?
> Did I provide all necessary information, don't I?

Yes.  thank you.  I haven't had a chance to look into this in depth yet.
Comment 9 ka.nick 2013-12-16 06:33:59 UTC
Mr. Deucher?
Is there any chance for a Christmas gift?
Comment 10 Alex Deucher 2013-12-21 14:44:16 UTC
This might be a duplicate of bug 60389.  Does attachment 90957 [details] [review] help?
Comment 11 Alex Deucher 2013-12-21 14:50:37 UTC
Created attachment 91088 [details] [review]
possible fix

You might also try this patch in combination with the previous one if it doesn't work.
Comment 12 ka.nick 2013-12-22 14:53:34 UTC
Unfortunately neither 1st one nor the combination of the two helped.
It just reshaped the garbage, and X still keeps restarting.
But now it looks like the screen is divided in small tiles that are maybe correct by themselves but placed at wrong positions and repeated now and then.
Comment 13 ka.nick 2013-12-22 15:15:33 UTC
UPD: Tried to set ColorTiling and ColorTiling2D to off.
In this case about 60% of the screen is rendered correctly but is placed @ wrong position on the screen. Mouse works but X restarts just the same.
Comment 14 Martin Peres 2019-11-19 08:39:52 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/amd/issues/401.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.