Bug 6573

Summary: X server freezes after a couple of seconds
Product: xorg Reporter: Andre Colomb <acolomb>
Component: Driver/RadeonAssignee: Xorg Project Team <xorg-team>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: high    
Version: 6.9.0   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
X server config file
none
X server log file none

Description Andre Colomb 2006-04-13 05:08:53 UTC 
When starting the Xorg server using the r300 driver on my IBM ThinkPad R52 
[ATI Mobility Radeon X300 M22 / PCI ID: 1002:5460] with debian 
testing/unstable (Xorg 6.9.0.dfsg.1-6, libgl1-mesa-dri 6.4.1-0.4, kernel 
2.6.16), it seems to come up alright even with 3D acceleration enabled.

After a couple of seconds of moving around the mouse, the cursor disappears 
and nothing moves anymore. Opening any window has the same effect, plus the 
contents don't get drawn completely (xterm is just a white rectangle with 
little pieces of the X background still showing on the bottom few pixels; 
icewmbg leaves an area of about 200x1 pixels on the bottom right corner 
uncovered). 

Neither server kill (Ctrl-Alt-BkSpace) nor VT switching work. The system is 
still usable over ssh, but trying to terminate the server freezes it 
completely (not even SysRq works). Server kill (Signal 9) works, but doesn't 
restore the console. Starting a framebuffer [radeonfb] X server afterwards 
also results in lockup.

strace over ssh shows me that the Xorg process is still responding to mouse 
movements (lots of select()s, read()s, gettimeofday()s) but it always tries to 

ioctl(9, 0x6444, 0) 

which returns with code -1, errno is EBUSY (Device or resource busy). File 
descriptor 9 refers to /dev/dri/card0. 

This seems to be in an endless loop, as the Xorg process consumes 99% of the 
CPU time. Maybe a race condition in the X server / r300 driver? I couldn't 
figure out what ioctl 0x6444 is.

I tried the closed-source fglrx driver from ATI, but it also freezes after an 
undetermined amount of time (especially when drawing lots of widgets--movies 
using Xv run for a couple of hours). For me, this indicates that it is a 
problem with the X server, as the ATI driver shows the same lockup symptoms 
and apparently a race condition. Windows XP runs without any troubles, so it 
is unlikely to be a hardware defect.

Note: It does not matter whether I enable the DynamicClocks option--this is 
not bug #2187!

Hope this helps improving the driver--Thanks!
Comment 1 Michel Dänzer 2006-04-13 16:02:05 UTC 
Please attach (as opposed to paste) full X config and log files.
Comment 2 Andre Colomb 2006-04-17 07:30:58 UTC 
Created attachment 5335 [details]
X server config file

The "Radeonserver" ServerLayout is the one I used for the test.
Comment 3 Andre Colomb 2006-04-17 07:33:41 UTC 
Created attachment 5336 [details]
X server log file

I'll also try to send the strace log as soon as I have another host up in the
network again.
Comment 4 Michel Dänzer 2006-04-17 21:08:07 UTC 
A lot of problems like this have been fixed in xf86-video-ati CVS. Would be
great if you could try either HEAD or ati-1-0-branch.
Comment 5 Michel Dänzer 2006-04-17 21:29:38 UTC 
Forgot to mention that not enabling the DRI might serve as a workaround.
Comment 6 Andre Colomb 2006-07-23 09:46:04 UTC 
I just came around to test this situation once again. What I am now using:
xserver-xorg 1:7.0.22, libgl1-mesa-dri 6.4.2-1, xserver-xorg-video-ati 1:6.5.8.0-1
from Debian testing and a self-compiled linux 2.6.17.6 kernel.

As suggested in some forums, I tried turning off the radeonfb driver in the
kernel. That solved the problem. However, I would still like to use it because
it gives me a 1400x1050 framebuffer console instead of a 1280x1024 with the
vesafb driver. Is that more of a kernel issue, rather than the X server?

Besides, I get this following error message when trying to run glxgears:

---
acolomb@pythagoras:~> glxgears 
Unknown device ID 5460, please report. Assuming plain R300.
*********************************WARN_ONCE*********************************
File r300_state.c function r300Enable line 456
TODO - double side stencil !
***************************************************************************
No ctx->FragmentProgram._Current!!
*********************************WARN_ONCE*********************************
File r300_render.c function r300_get_num_verts line 188
user error: Need more than 2 vertices to draw primitive QS !
***************************************************************************
---

lspci correctly identifies my card as an ATI Technologies Inc M22 [Radeon
Mobility M300], so I guess the PCI ID just has to be added to the driver. The
rest of the messages is probably due to the fact that the driver is still
incomplete (Correct me if I'm wrong...)

Hope this information helps improving the driver further. For now I can use the
radeon driver in my default X screen and watch hardware-scaled movies :-) Would
be really nice to use the kernel radeonfb driver, though.
Comment 7 Timo Jyrinki 2007-02-22 14:28:13 UTC 
Marking broken (status null/blank) bugs in xorg with no activity in a long time as fixed. Please reopen if you think it's necessary, but first do a search if a similar bug report is already filed and in a NEW/ASSIGNED state. These bugs do not currently show in most search results as they do not have any status.

Sorry for this janitorial spam, you know where to send hate mails to when your inbox gets full of bugs you're subscribed to.

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.