Bug 97578 - [EXA] Radeon HD 5450 suddenly blanks screen or shows old content
Summary: [EXA] Radeon HD 5450 suddenly blanks screen or shows old content
Status: RESOLVED MOVED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/Radeon (show other bugs)
Version: 7.7 (2012.06)
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: xf86-video-ati maintainers
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-09-02 17:30 UTC by Markus Näher
Modified: 2019-11-19 07:57 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Xorg.0.log (65.71 KB, text/plain)
2016-09-07 15:49 UTC, Markus Näher
no flags Details
Radeon HD 5450 corrupted display (109.07 KB, image/png)
2016-09-08 15:56 UTC, Markus Näher
no flags Details
Radeon HD 5450 correct display (1.82 MB, image/png)
2016-09-08 15:59 UTC, Markus Näher
no flags Details
XFCE compositor settings (60.07 KB, image/png)
2016-09-14 08:27 UTC, Markus Näher
no flags Details

Description Markus Näher 2016-09-02 17:30:44 UTC
Some Background info:

I'm using the radeon driver for my Radeon HD 5450 on ArchLinux (x86_64) with XFCE desktop.

The issue occurs since one of the latest software updates, on two PC's with a Radon HD 5450, both with two screens attached. Because of the two PC's, I think a hardware defect is unlikely. On one of the PC's, I've tried a Radeon HD 7750 (borroewd from a friend), and the issue was gone until I had to give it back. I did not change any settings during the test.

This is what happens:

Sometimes, it "forgets" what it should display on the screen and suddenly, I see a mix of an old screen content, with some black areas and (most times, but not always) a part of the xscreensaver dialog. The issue can be triggered by opening a window, menu, or dialog or even some animation within a window. One thing that often "works" as a trigger is the "blue arrow" animation of firefox when I start a download. Sometimes it can just be the blinking of the cursor in a text editor like geany. But the trigger is not always that obvious.

When I try to force a redraw of the whole screen by switching between virtual desktops or between console and X, it shows the correct screen content for a split second, and then switches to the wrong display again. It looks like flickering between two states.

The only way (other than rebooting) I found out to stop it is to identify the trigger window and close (or hide) it. Sometimes I can restore it without triggering again, sometimes not. Sometimes it stops after a few minutes, sometimes not. Until it stops, the PC is completely unusable. But even identifying the right window and trying to hit the minimize or close button with the mouse when you can't see what you're doing is really hard.

As of the PC's has long-running jobs that can not be suspended and resumed, rebooting is mostly not an option.

I'm not 100% sure, but I think it happens after the lockscreen once was activated, either manually or by idle timer. From reboot to first lockscreen, I think I haven't experienced it yet. Other than lockscreen, I don't use any screensaver.

The screen content that is displays can be really old: Yesterday, I changed my desktop background, and today, I saw the old background when it happened (no reboot in between).


I really would like to keep my 5450's, because they are really good office cards (passive cooling, dual DVI-I, ...) and even good enough for some gaming. But the issue literally drives me crazy.
Comment 1 Tom St Denis 2016-09-02 17:54:09 UTC
Just so happens I have an MSI 5450 in my workstation (long story short the ball bearings in my Bonaire gave out and I wanted something quiet for text editing...).

I'm using the latest upstream UMD and KMD from Fedora 24 without issue since day 1.  

This is very likely a distro issue.
Comment 2 Markus Näher 2016-09-02 23:01:57 UTC
OK. Let's try to find out if it's a distro, XFCE, or xscreensaver issue.
Are you also using XFCE and xscreensaver ?
Comment 3 Michel Dänzer 2016-09-05 09:29:34 UTC
Tom, let's not jump to conclusions.

Markus, please attach the corresponding Xorg log file.

(In reply to Markus Näher from comment #0)
> The issue occurs since one of the latest software updates, [...]

Which packages were updated as part of that?
Comment 4 Michel Dänzer 2016-09-07 14:14:16 UTC
Markus, did you see comment 3? The information I asked for there is important for making progress on this report.
Comment 5 Markus Näher 2016-09-07 15:47:57 UTC
The packages that have been upgdaded at the moment when the issue occurred irst are hard to say, since it's been a few months since then.
The reason why I reported the bug so late was that I first tried to get help from the XFCE folks but then they said it looks like a driver issue.
Second reason is that I used the borrowed Radeon HD 7750 on the more important PC for quite a while, hoping that the bug would be found and fixed before I need to give it back.

I'm attaching my Xorg.log, but it show only the massages from startup. I've just double-checked that there is no log when the issue occurs: I rebooted the PC two days ago, and on the day after reboot, the isue occcurred the first time. After that, the file content and date were still the same as after reboot.

I'm trying to create some screenshots. But when I hit the "print" key (associated with xfce4-screenshooter), it immediately redraws parts of the screen and the redrawed screen is then saved as screenshot, and after that it flickers back to the corrupted display.
Comment 6 Markus Näher 2016-09-07 15:49:27 UTC
Created attachment 126284 [details]
Xorg.0.log
Comment 7 Tom St Denis 2016-09-07 15:56:48 UTC
FWIW Fedora 24 stable has X 1.18.3 and xf86-ati-driver 7.6.99 according to the log.

So could very well be something new in the X or radeon usermode driver.
Comment 8 Michel Dänzer 2016-09-08 06:32:31 UTC
Markus, is there any chance you can try if the problem also occurs with current upstream xf86-video-ati Git master, or alternatively with the xf86-video-ati 7.6.0 release?

Does turning on/off the XFCE compositor avoid the problem? If so, are there any specific XFCE compositor settings with an effect on the problem?
Comment 9 Markus Näher 2016-09-08 15:54:30 UTC
Michael, yes, it looks like there is a package named xf86-video-ati-git for my distribution. It's currently flagged "out of date", but I can try and check what exactly is out of date and then test it. Before that, I would like to set up a backup system on a second partition, just in case I screw up something. Could take a few days ...

Tomorrow, I can test if turning off compositing makes a difference.

Today, I managed to create a screenshot that shows the effect very well. I blurred some parts for privacy :-)
I'm attaching two screenshots: one when the screen was corrupted and a second one shows how it should have looked like in that moment. This time I was lucky and could "fix" it by switching back and forth between virtual desktops in order to redraw the whole screen. After this, everything worked for an hour or so.

There is one more interesting thing: When creating a screenshot, the current screen content becomes the new "saved state" that always is displayed when the issue occurs. On top of that, a varying number of window fragments are drawn. E.g. the CPU usage graph, sometimes the clock etc ...

BTW ... when the screen is corrupted, x11vnc also shows the corrupted content. No, x11vnc is not running all the time, and the issue also occurs when x11vnc is stopped. :-)
Comment 10 Markus Näher 2016-09-08 15:56:16 UTC
Created attachment 126309 [details]
Radeon HD 5450 corrupted display
Comment 11 Markus Näher 2016-09-08 15:59:21 UTC
Created attachment 126310 [details]
Radeon HD 5450 correct display
Comment 12 Michel Dänzer 2016-09-09 00:29:14 UTC
Something else you could try is whether

 Option "AccelMethod" "glamor"

in /etc/X11/xorg.conf has any effect on the problem.
Comment 13 Markus Näher 2016-09-09 12:32:43 UTC
How do I set it in xorg.conf.d ?
I just checked the docs at https://www.x.org/archive/current/doc/man/man5/xorg.conf.5.xhtml, but could not find anything about it.

Disabling compositing seems to help. No occurrences so far today.
Comment 14 Michel Dänzer 2016-09-12 01:25:18 UTC
(In reply to Markus Näher from comment #13)
> How do I set it in xorg.conf.d ?

You don't. :) xorg.conf.d is for system configuration snippets, users should always use /etc/X11/xorg.conf directly. Something like this should do:

Section "Device"
        Identifier      "default"
        Driver          "radeon"
        Option          "AccelMethod" "glamor"
EndSection


> Disabling compositing seems to help. No occurrences so far today.

Can you attach screenshots of all XFCE settings tabs containing any compositing related settings?
Comment 15 Markus Näher 2016-09-14 08:26:03 UTC
Well ... I asked about xorg.conf.d because there is no xorg.conf on my PC's. As far as I understand, this means my X server is autoconfigured except a few things in my xorg.conf.d. Currenntly, it's only keyboard settings and dual screen positions (DVI-1 right of DVI-0).

I'm trying your proposed settings as soon as my current long running job is done. It runs in a GUI app, so I can't even restart X.

About the radeon-git package ... as I said, it is flagged as outdated, but I found out it's even worse ... the maintainer abandoned it. So I'm afraid I won't be able to install the git version.
But I have another idea: I was lucky to find a cheap used Radeon HD 7750 with passive cooling, so one of my two PC's does not have the problem anymore, and one of the 5450's is now left over.
As your name looks like it could be german (and hopefully you live in europe), I could send you the left 5450 via mail.
This could be easier (for both of us) than you advising me how to test and possibly debug.

If that's OK for you, you can contact me via the EMail link on my name and give me your postal address.

I promise not to bother you :-)

Another interesting thing; When I re-enabled the compositor for making the screenshot of the settings, the issue showed up immediately.
Comment 16 Markus Näher 2016-09-14 08:27:15 UTC
Created attachment 126510 [details]
XFCE compositor settings
Comment 17 Michel Dänzer 2016-09-14 09:01:11 UTC
(In reply to Markus Näher from comment #15)
> Well ... I asked about xorg.conf.d because there is no xorg.conf on my PC's.
> As far as I understand, this means my X server is autoconfigured [...]

Yep, that doesn't change what I said. :)


> As your name looks like it could be german (and hopefully you live in
> europe), I could send you the left 5450 via mail.
> This could be easier (for both of us) than you advising me how to test and
> possibly debug.

My last name is German (though I'm Swiss), but I'm currently living outside of Europe. Anyway, I do have a 5xxx card, the problem is "just" finding the time to try and reproduce and fix the problem; I'm afraid it's not very high priority for me.

Anyway, since the problem doesn't happen with a 7750, it's quite possible that using glamor will avoid it with a 5450 as well.
Comment 18 Markus Näher 2016-09-16 11:42:31 UTC
Today, I activated glamor and re-enabled compositing.

Additionally, I did my routinely system upgrades, and one of the upgraded packages was xf86-video-ati from 1:7.7.0-1 to 1:7.7.1-1.

Testing is in progress, and looks good so far. End of next week, I will report again if it's really totally gone. I would like to test a whole week because I always had the impression that it's getting worse with growing uptime.
Comment 19 Markus Näher 2016-09-25 07:05:52 UTC
It did not happen any more with glamor.

Thank you for your help.
Comment 20 Michel Dänzer 2016-09-27 08:52:23 UTC
(In reply to Markus Näher from comment #19)
> It did not happen any more with glamor.

Does it still happen with EXA with 7.7.1?
Comment 21 Markus Näher 2016-10-10 10:03:05 UTC
Yes, it does. Occurred 1 minute ago.

Sorry for the delay, but I try to reboot (or restat X) as rarely as possible.
Comment 22 Markus Näher 2016-10-10 10:32:33 UTC
There is one more thing I'd like to ask ...

Would it improve performance if I configure glamor for the PC running with the Radeon HD 7750 also ?

And even more important: I have set up a multimedia PC for my parents with an on-board Radeon HD 4200. It has sometimes performance problems playing HD-video. Would glamor help there ?
Comment 23 Alex Deucher 2016-10-10 14:27:11 UTC
(In reply to Markus Näher from comment #22)
> There is one more thing I'd like to ask ...
> 
> Would it improve performance if I configure glamor for the PC running with
> the Radeon HD 7750 also ?

SI and newer only support glamor. So that is the only option.

> 
> And even more important: I have set up a multimedia PC for my parents with
> an on-board Radeon HD 4200. It has sometimes performance problems playing
> HD-video. Would glamor help there ?

Possibly.  More important would be to make sure vdpau is enabled properly for hw accelerated video decoding and that the app you are using for playback is properly configured.
Comment 24 Martin Peres 2019-11-19 07:57:30 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/xorg/driver/xf86-video-ati/issues/163.


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.