Bug 38382 - radeon: wait for empty RBBM fifo failed
Summary: radeon: wait for empty RBBM fifo failed
Status: RESOLVED DOCUMENTED
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-06-28 09:09 UTC by Stefano Avallone
Modified: 2012-08-24 15:02 UTC (History)
2 users (show)

See Also:
Kernel Version: 3.0.0-rc4
Subsystem:
Regression: No
Bisected commit-id:


Attachments
dmesg with pageflipping disabled (53.54 KB, text/plain)
2011-06-28 09:09 UTC, Stefano Avallone
Details
Xorg log with pageflipping disabled (41.30 KB, text/plain)
2011-06-28 09:10 UTC, Stefano Avallone
Details

Description Stefano Avallone 2011-06-28 09:09:35 UTC
Created attachment 63662 [details]
dmesg with pageflipping disabled

My system is affected by frequent oopses that makes it unusable. There are no problems when the X server is not launched, but when it starts (with KDM) something bad soon happens. If I don't login into KDM, it may freeze after few minutes. If I login, an oops happen after few minutes. Unfortunately I can't tell exactly when this behavior started to happen, but somewhere between 2.6.39 and 2.6.39.1. Yesterday I compiled the kernel from git but no luck. I tried to disable pageflipping (as suggested in bug #32402) but the oops happened again. WIth pageflipping disabled, I managed to ssh into the machine and get dmesg and Xorg.log (attached). Unfortunately, I didn't manage to ssh into the machine with pageflipping enabled.

I have a Radeon X300 card and I am using mesa git from 20110616, libdrm 2.4.25 and Xorg 1.10.2.

If I need to test anything, please let me know, thanks.
Comment 1 Stefano Avallone 2011-06-28 09:10:54 UTC
Created attachment 63672 [details]
Xorg log with pageflipping disabled
Comment 2 Stefano Avallone 2011-06-28 09:12:34 UTC
Sorry, forgot to say that I am using the r300g gallium driver.
Comment 3 Michel Dänzer 2011-06-28 09:18:31 UTC
(In reply to comment #0)
> Unfortunately I can't tell exactly when this behavior started to happen, but
> somewhere between 2.6.39 and 2.6.39.1.

Can you bisect?
Comment 4 Stefano Avallone 2011-06-28 10:04:37 UTC
Well, I re-installed kernel 2.6.39 and even 2.6.38.6, but I still get the same problems. The last times I tried, the system freezed for a few seconds, then the monitor got powered off and I was unable to ssh into the system. 

I know that I have provided too few information. I don't remember well, but maybe I started to got these problems without updating any software (though I am not 100% sure). Might it be an hardware problem of the video card?

Any suggestion for possible tests is very appreciated, thanks.
Comment 5 Michel Dänzer 2011-06-28 10:20:10 UTC
(In reply to comment #4)
> I know that I have provided too few information. I don't remember well, but
> maybe I started to got these problems without updating any software (though I
> am not 100% sure). Might it be an hardware problem of the video card?

It's possible, but more likely the problems started after changing other software components, e.g. Mesa or X (e.g. the video driver).
Comment 6 Stefano Avallone 2011-07-29 14:56:55 UTC
I re-booted my PC under Windows, which shows me messages like

"VPU recover has reset your graphics accelerator as it was no longer responding to graphics driver commands"

and 

"VPU recover was unable to fully recover from a hardware deadlock and has switched to software rendering. To restore hardware rendering you must restart your computer. Do you want to restart now?"

After re-booting, the same messages appear. If booting under Linux, the PC shuts down a few seconds (variable) after logging into KDE.

Any idea if there is a solution to this problem? Or should I simply buy a new graphics card?

Thanks.
Comment 7 Alex Deucher 2011-07-29 15:04:29 UTC
Sounds like the card is going bad.  You might check to make sure the heat sink is properly seated on the asic and vram and the fan is working a free of dust.

Note You need to log in before you can comment on or make changes to this bug.