Bug 45641

Summary: Screen goes black randomly
Product: DRI Reporter: John <john.ettedgui>
Component: DRM/RadeonAssignee: Default DRI bug account <dri-devel>
Status: CLOSED NOTABUG QA Contact:
Severity: normal    
Priority: medium CC: honyczek, wjones
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Xorg log
none
dmesg
none
possible fix none

Description John 2012-02-05 02:31:59 UTC
Hello,

First thing I am not sure if this is driver related or not, but I believe so hence this bug.

So since around July or so (linux 3.0), my screen has been randomly going black (like when it is in power saving mode).

Moving the mouse seems to bring it back faster.
I do not believe it to be a screen saver normal behavior as it can go black even when I am using the mouse.

On the other hand I never had that issue in XBMC fullscreen (or in Windows).

I am running linux 3.2 (but it has been the same since 3.0) KDE SC, now 4.8, and these from git:


dri2proto-git 20120129-1
glproto-git 20120129-1
lib32-libdrm-git 20120129-1
lib32-mesa-full 20120129-1
libdrm-git 20120129-1
mesa-full 20120129-1
xf86-video-ati-git 20120129-1


I have a RadeonHd 4670.

Not sure what else I can provide.

Thank you,
John
Comment 1 Alex Deucher 2012-02-06 06:49:03 UTC
Please attach your xorg log and dmesg output.
Comment 2 John 2012-02-06 07:23:15 UTC
Created attachment 56676 [details] [review]
Xorg log
Comment 3 John 2012-02-06 07:24:06 UTC
Created attachment 56677 [details] [review]
dmesg
Comment 4 John 2012-02-06 07:24:35 UTC
There you.
Thanks
Comment 5 Alex Deucher 2012-02-07 10:13:49 UTC
Does the blanking happen are regular intervals (e.g., every 30 seconds, every 2 minutes, etc.)?  If it worked with 3.0, can you bisect the kernel to see what commit is causing the problem?
Comment 6 John 2012-02-07 12:22:13 UTC
It seems quite random to me.
Also I remember it happening around the 3.0 release, but I cannot say that the issue is coming from there.
For all I know it could be KDE messing up somewhere or something like that, since it does not happen with XMBC fullscreen.

Because of that, and the fact that the problem is quite random, I don't know if it  is best to bitsect the kernel already.

Do you know of any (debug) log I can find that could point me at the culprit a bit?

Thanks!
Comment 7 Warren Jones 2012-06-15 16:36:50 UTC
I'm experiencing similar problems.  My screen will go black for 1-2 seconds at random intervals, typically more than a dozen times per hour.  In a two monitor configuration, sometimes the left monitor will go black, sometimes the right monitor, and occasionally both.

My video card:  ATI RV610 [Radeon HD 2400 XT]

My configuration:

    openSUSE 12.2 beta 1
    kernel-desktop-3.4.0-2.2
    xorg-x11-7.6_1-1.1
    xf86-video-ati-6.14.4-2.1

The problem appeared somewhere between openSUSE 12.2 milestone 3 and beta 1.  This is the last known good configuration:

    openSUSE 12.2 milestone 3
    kernel-default-3.3.0-2.1
    xorg-x11-7.6-76.3
    xorg-x11-driver-video-7.6-91.1

(The radeon driver moved from xorg-x11-driver-video to xf86-video-ati between these two releases.)

eg> cat /etc/X11/xorg.conf.d/50-device.conf
Section "Device"
  Identifier   "Default Device"
  VendorName   "ATI"
  BoardName    "Radeon HD 2400 XT"
  Driver       "radeon"
  Option       "monitor-DVI-0" "Monitor[1]"
  Option       "monitor-DVI-1" "Monitor[0]"
EndSection

I installed kernel-default-3.3.0-2.1 from the last known good configuration to rule out problems related to KMS, but the black screens continued.

I've observed the problem with KDE, Gnome, XFCE, LXDE, etc.

I don't see anything suspicious in /var/log/Xorg.0.log, /var/log/messages or the output from dmesg.

I've also filed a bug report here:

    https://bugzilla.novell.com/show_bug.cgi?id=767360
Comment 8 Jan Papež (honyczek) 2013-03-19 10:16:46 UTC
The issue still remains.

openSUSE 12.3
kernel-desktop  3.7.10-1.1.1
xorg-x11-server 7.6_1.13.2-1.2.1
xf86-video-ati  7.0.0-2.1.1
libdrm2         2.4.42-1.1.1
libdrm_radeon1  2.4.42-1.1.1
Mesa            9.0.2-34.3.1

I have ATI Radeon HD 2400 XT (RV610).
Comment 9 Alex Deucher 2013-03-19 12:53:14 UTC
Does booting with radeon.disp_priority=2 on the kernel command line in grub help?  For systems where this is a regression, can you bisect?
Comment 10 Jan Papež (honyczek) 2013-03-19 13:43:07 UTC
> Does booting with radeon.disp_priority=2 on the kernel command line in grub help?

No.

> For systems where this is a regression, can you bisect?

I don't understand. Could you explain this question please (if it was to me)?
Comment 11 Alex Deucher 2013-03-19 14:01:23 UTC
(In reply to comment #10)
> > For systems where this is a regression, can you bisect?
> 
> I don't understand. Could you explain this question please (if it was to me)?

If it worked with an older kernel, you can use git to bisect between working kernel and the the broken one.  Bisecting allows you to pinpoint the change that caused the problem.  There are a number of good tutorials on using git to bisect a problem.
Comment 12 Jan Papež (honyczek) 2013-03-19 14:06:26 UTC
Additional specific info:
I'm using extended desktop on two DVI connected monitors (the card has one connector and you have to use Y cable to connect something). Even if I switch to cloned desktop, it blinks too.

Interesting: If I connect only one monitor (either on cable 1 or 2), it works well.

Both monitors have resolution 1680x1050.

My ideal monitor position would be desktop assembled from 1050x1680 (rotated left) and 1680x1050.

With fglrx it was good working...
Comment 13 Alex Deucher 2013-03-19 14:22:41 UTC
(In reply to comment #12)
> Interesting: If I connect only one monitor (either on cable 1 or 2), it
> works well.

The blinking is caused by data underflow to the display controllers.  All clients on the GPU (displays, 3D engine, etc.) contend for memory bandwidth.  If you have a card with limited memory bandwidth (like yours), you may get underflow if there is not enough bandwidth to service the displays at a particular moment in time.  Running several large monitors increases the memory bandwidth requirements.  The display watermarks may need some fine tuning for situations like yours.  Additional things to try:
1. try a lower refresh rate
2. Disable acceleration (Option "NoAccel" "True" in the device section of your xorg.conf)
3. Disable Tiling (Option "ColorTiling" "False" in the device section of your xorg.conf)
4. Try mesa 9.1 or newer.
Comment 14 Jan Papež (honyczek) 2013-03-19 14:49:05 UTC
(In reply to comment #13)

> 2. Disable acceleration (Option "NoAccel" "True" in the device section of your xorg.conf)
> 3. Disable Tiling (Option "ColorTiling" "False" in the device section of your xorg.conf)

Is possible to add these options when using KMS? Modifying /etc/X11/xorg.conf.d/*.conf is enough?
Comment 15 Alex Deucher 2013-03-19 16:35:38 UTC
(In reply to comment #14)
> Is possible to add these options when using KMS? Modifying
> /etc/X11/xorg.conf.d/*.conf is enough?

yes.
Comment 16 Jan Papež (honyczek) 2013-03-22 12:24:30 UTC
(Additional info to comment #12)
> Interesting: If I connect only one monitor (either on cable 1 or 2), it works well.
It is not working well, but in only blinks in longer intervals. After resuming from hibernation, it started blinking more.
(In reply to comment #13)
> 2. Disable acceleration (Option "NoAccel" "True" in the device section of your xorg.conf)

After starting Xorg with NoAccel option, it is not blinking. But:
1. why don't use acceleration on accelerated graphic card?
2. I can't use rotated monitor, because it requires acceleration enabled.
Comment 17 Alex Deucher 2013-03-22 14:15:40 UTC
Created attachment 76907 [details] [review]
possible fix

Does the attached kernel patch help?
Comment 18 John 2013-04-01 01:05:04 UTC
I tried the patch today but got another black screen already.
I have an 4670 by the way, not sure if the patch was made for that one or not...

Thanks
Comment 19 John 2013-10-03 17:57:35 UTC
Still same behavior with current drivers/kernel/mesa, I was hoping something would have helped :)

Alex, do you have any other patch I can test against this?

Thanks,
John

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.