FS#19092 - [xorg18] [kms] [xf86-video-ati] make your choice: crashes or glitches

Attached to Project: Arch Linux
Opened by vnoel (kaouete) - Wednesday, 14 April 2010, 13:30 GMT
Last edited by Andreas Radke (AndyRTR) - Monday, 21 June 2010, 19:50 GMT
Task Type Bug Report
Category Packages: Extra
Status Closed
Assigned To Jan de Groot (JGC)
Andreas Radke (AndyRTR)
Architecture x86_64
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 2
Private No

Details

Hello,

I am using the xorg18 repository, core and extra (no testing, but I checked, there should be nothing important in it currently).
I have a ATI Mobility Radeon HD 3450.

There is two problematic situations to report:
1) When kms is enabled (using 2.6.33+radeon_ucode+radeon-initrd+earlystart), all is working well except that X did crash when I was playing with blender, and I remember that every time I did try kms, X crashed in one way or another (even when not playing with 3D apps).
There is error like that in dmesg:
[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !

2) When kms is disabled, all is working except that X shows a lot of graphical glitches that makes it hard to use, it is often when scrolling.
There is errors like that in dmesg:
[drm] wait idle failed status : 0xE77704E0 0x00110103
...
[drm] wait for fifo failed status : 0xE77304E0 0x00110103

Tell me if you need more informations or if I should report this somewhere else…
This task depends upon

Closed by  Andreas Radke (AndyRTR)
Monday, 21 June 2010, 19:50 GMT
Reason for closing:  Fixed
Comment by vnoel (kaouete) - Wednesday, 14 April 2010, 16:25 GMT
Note: by "crash" I mean that X freezes and that the computer doesn't answer at all (not even by ssh).
Comment by Andreas Radke (AndyRTR) - Wednesday, 14 April 2010, 16:44 GMT
please try to boot with kernel option "pci=nomsi".
Comment by vnoel (kaouete) - Wednesday, 14 April 2010, 17:47 GMT
I guess it is for when KMS is enabled since it doesn't remove the glitches :)

I will try with this option on and will see what happens, thanks.
Comment by vnoel (kaouete) - Wednesday, 14 April 2010, 18:00 GMT
It froze even faster!

I just started to do things with blender when my right screen (the one on which blender was started) turned black and the left screen (the one on my laptop) was streaked with white lines.
At first I could see the drawing behind the streak but then it evolved and at some time (few seconds) it was blank…

Last message I can see in everything.log before next boot is:
Apr 14 19:47:46 volte kernel: [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !

Last messages I get in Xorg.0.log are:
[ 83.323] (II) Power Button: Close
[ 83.323] (II) UnloadModule: "evdev"
[ 83.333] (II) Video Bus: Close
[ 83.334] (II) UnloadModule: "evdev"
[ 83.344] (II) Sleep Button: Close
[ 83.344] (II) UnloadModule: "evdev"
[ 83.353] (II) TypeMatrix.com USB Keyboard: Close
[ 83.353] (II) UnloadModule: "evdev"
[ 83.364] (II) TypeMatrix.com USB Keyboard: Close
[ 83.364] (II) UnloadModule: "evdev"
[ 83.384] (II) Logitech Optical USB Mouse: Close
[ 83.384] (II) UnloadModule: "evdev"
[ 83.397] (II) CKA7216: Close
[ 83.397] (II) UnloadModule: "evdev"
[ 83.410] (II) AT Translated Set 2 keyboard: Close
[ 83.410] (II) UnloadModule: "evdev"
[ 83.444] (II) UnloadModule: "synaptics"
[ 83.464] (II) PS/2 Generic Mouse: Close
[ 83.464] (II) UnloadModule: "evdev"
[ 83.484] (II) Macintosh mouse button emulation: Close
[ 83.484] (II) UnloadModule: "evdev"
[ 83.497] (II) HP WMI hotkeys: Close
[ 83.497] (II) UnloadModule: "evdev"
[ 83.511] Static power management disable success
[ 83.511] (II) RADEON(0): Dynamic Clock Gating Disabled
[ 83.512] (II) RADEON(0): Power Mode Switch
[ 84.092] Output DIG0 transmitter setup success
[ 84.095] Blank CRTC 0 success
[ 84.096] Disable CRTC memreq 0 success
[ 84.112] Disable CRTC 0 success
[ 84.112] Output DIG0 transmitter setup success
[ 84.123] Blank CRTC 1 success
[ 84.123] Disable CRTC memreq 1 success
[ 84.140] Disable CRTC 1 success
[ 84.148] (II) RADEON(0): RADEONRestoreMemMapRegisters() :
[ 84.148] (II) RADEON(0): MC_FB_LOCATION : 0x00c700c0 0x00c700c0
[ 84.148] (II) RADEON(0): MC_AGP_LOCATION : 0x00000000
[ 84.157] (II) RADEON(0): avivo_restore !
[ 84.177] Enable CRTC 0 success
[ 84.177] Enable CRTC memreq 0 success
[ 84.190] Unblank CRTC 0 success
[ 84.204] (II) RADEON(0): [drm] removed 1 reserved context for kernel
[ 84.204] (II) RADEON(0): [drm] unmapping 8192 bytes of SAREA 0x2b7ff000 at 0x7f72f04e3000
[ 84.205] (II) RADEON(0): [drm] Closed DRM master.

Theses messages are not in the Xorg.0.log when running without KMS…

If you have other ideas for messages to gather…
Comment by Andreas Radke (AndyRTR) - Wednesday, 14 April 2010, 18:13 GMT
What was the state with KMS/UMS and Xorg from extra? Is your issue Xorg/mesa related or kernel-drm related? Have you tried more recent drm module (rc kernel or drm-next) or a driver git shot? Maybe you should ask the upstream devs if this is/was a known issue.
Comment by vnoel (kaouete) - Thursday, 15 April 2010, 12:29 GMT
I will try today with kms and xorg18 until it freeze, but without using a 3D app. Not sure it will happen.

Then I will try with KMS/UMS and Xorg from extra to see if it freeze with blender, if it is the case I guess we could say that it is kernel related, but since I just enabled and did not used it with 2.6.32 we won't be sure…

Maybe I should use the git version of the kernel…

I will see with upstream for the follow-ups and report here if I have more information.

Thanks or your help, if you have more ideas to test, do not hesitate :)
Comment by vnoel (kaouete) - Thursday, 15 April 2010, 13:19 GMT
Just for more information: it doesn't freeze with blender and current kms/ums/xorg from extra.

So it seems to be linked to the package available in xorg18, if I understand right, it means that it is Xorg/mesa related?
Comment by vnoel (kaouete) - Friday, 16 April 2010, 12:57 GMT Comment by Andreas Radke (AndyRTR) - Monday, 21 June 2010, 12:23 GMT
Is this solved in Xorg18/kernel 2.6.34?
Comment by vnoel (kaouete) - Monday, 21 June 2010, 17:02 GMT
Hello,

I did not had the time to test it with KMS, but first, please notice that the bug reports was about xorg18.

Now, with current core+extra, so with xorg 1.8.1 and 2.6.34, without KMS I still get the glitches.

I will try with KMS.

Thanks
Comment by vnoel (kaouete) - Monday, 21 June 2010, 17:34 GMT
For now, it seems to work with KMS, Blender does not crash.

I just noticed some updates are already available (1.8.1.902 and mesa 7.8.2) so I will test without KMS later to see if there is a difference.
Comment by Jan de Groot (JGC) - Monday, 21 June 2010, 17:39 GMT
I don't know the roadmap for xf86-video-ati, but note that KMS is the future, UMS will likely get removed in the future.
Comment by Andreas Radke (AndyRTR) - Monday, 21 June 2010, 19:50 GMT
Because KMS is now the way to go I will close this one as fixed. If you will see crashes with KMS please open a new bug.

Loading...