FS#17357 - [xf86-video-ati] drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream.

Attached to Project: Arch Linux
Opened by Florian (monsieur.moche) - Thursday, 03 December 2009, 06:10 GMT
Last edited by Andreas Radke (AndyRTR) - Thursday, 24 December 2009, 09:08 GMT
Task Type Bug Report
Category Upstream Bugs
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 1
Private No

Details

Description:
I can't run neither Blender nor glxgears.

florian@flobox:~$ lspci | grep VGA
01:00.0 VGA compatible controller: ATI Technologies Inc RV710 [Radeon HD 4550]
florian@flobox:~$ glxinfo | grep 'dir\|rer'
direct rendering: Yes
OpenGL renderer string: Mesa DRI R600 (RV710 9540) 20090101 TCL
florian@flobox:~$ glxgears
drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.
florian@flobox:~$ dmesg |tail
[drm] Resetting GPU
[drm] Setting GART location based on new memory map
[drm] Loading RV710 PFP Microcode
[drm] Loading RV710 CP Microcode
[drm] Resetting GPU
[drm] writeback test succeeded in 1 usecs

I could run them before video-ati was updated from a stable version to a GIT one (which btw sometimes has rendering issues when scrolling in e.g. Firefox).
The bug still occurs with aur/xf86-video-ati-git.

Looking for info on the Internet, I saw a Ubuntu bug report which seems similar and which has been fixed: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/421245

Additional info:
* package version(s)
extra/libdrm 2.4.15-1
extra/mesa 7.6-2
extra/libgl 7.6-2
extra/ati-dri 7.6-2
extra/xf86-video-ati 6.12.99.git20091014-1
extra/xf86-video-vesa 2.2.1-1
extra/xorg-server 1.7.2-2
extra/xf86-input-evdev 2.3.1-1

* config and/or log files etc.
- no xorg.conf
- no special instruction in grub command
This task depends upon

Closed by  Andreas Radke (AndyRTR)
Thursday, 24 December 2009, 09:08 GMT
Reason for closing:  Fixed
Additional comments about closing:  pkg downgraded, KMS support has been dropped
Comment by Andreas Radke (AndyRTR) - Thursday, 03 December 2009, 18:01 GMT
What a poor documentation in the Ubuntu tracker :(

When it is still happening with a more recent git shot you should wait for kernel 2.6.32 that will have huge drm changes. It should appear in a few days in our testing repos.
Comment by Florian (monsieur.moche) - Thursday, 03 December 2009, 19:55 GMT
That's a good news!
OK so let's wait for the kernel update and I'll keep you informed.
Comment by Andreas Radke (AndyRTR) - Monday, 07 December 2009, 10:01 GMT
state? also new ati/libdrm packages are probably coming later today to testing repo. give it a try.
Comment by Florian (monsieur.moche) - Monday, 07 December 2009, 16:29 GMT
florian@flobox:~$ yaourt -Qs xf86-video-ati
testing/xf86-video-ati 6.12.99.git20091207-1 (xorg-video-drivers)
X.org ati video driver
florian@flobox:~$ yaourt -Qs libdrm
testing/libdrm 2.4.16-1
Userspace interface to kernel DRM services

The bug still occurs with these new packages.
Comment by VuDu (VuDu) - Friday, 11 December 2009, 00:04 GMT
I'm also having thins issue... I'll looking forward for a fix.
Comment by VuDu (VuDu) - Saturday, 12 December 2009, 15:39 GMT
I forgot to mention that I don't have this issue if I boot to kernel26-lts.
Comment by Florian (monsieur.moche) - Sunday, 13 December 2009, 15:27 GMT
I just upgraded kernel26 package from the testing repo:
florian@flobox:~$ uname -a
Linux flobox 2.6.32-ARCH #7 SMP PREEMPT Fri Dec 4 15:39:16 CET 2009 x86_64 Intel(R) Core(TM) i5 CPU 750 @ 2.67GHz GenuineIntel GNU/Linux

Glxgears still crashes, but the error message is slightly different:
florian@flobox:~$ glxgears
IRQ's not enabled, falling back to busy waits: 2 0
drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.
florian@flobox:~$ dmesg |tail
[drm:r600_cs_packet_next_reloc_nomm] *ERROR* No packet3 for relocation for packet at 45.
[drm:r600_packet3_check] *ERROR* bad SET_CONTEXT_REG 0x28014
[drm:r600_cs_legacy] *ERROR* Invalid command stream !
Comment by Mihai Militaru (akephalos) - Wednesday, 23 December 2009, 01:59 GMT
I subscribe to this bug. x86_64, Radeon HD 4570, I tried both kernels - 2.6.31 current and 2.6.32 from testing some days ago.
Comment by Andreas Radke (AndyRTR) - Wednesday, 23 December 2009, 11:42 GMT
Please check again with packages from Core/Extra. We dropped all the experimental KMS/DRI2 code. Expect new cards to fall back to swrast being slow again but working. See discussion on the dev public list.
Comment by Florian (monsieur.moche) - Wednesday, 23 December 2009, 21:34 GMT
Great, both glxgears and Blender work again!
This isn't perfect yet though.


Glxgear raises a fatal error when closing the window.
This is sometimes a 11 error and sometimes a 104 error:

florian@flobox:~$ glxgears
IRQ's not enabled, falling back to busy waits: 2 0
XIO: fatal IO error 104 (Connection reset by peer) on X server ":0.0"
after 14663 requests (43 known processed) with 0 events remaining.
florian@flobox:~$ glxgears
IRQ's not enabled, falling back to busy waits: 2 0
XIO: fatal IO error 11 (Resource temporarily unavailable) on X server ":0.0"
after 10058 requests (43 known processed) with 0 events remaining.

There's no such an error when quitting Blender.


Another issue: directly rendered graphics are always on top of the undirectly rendered ones (see attached screenshot). It happens with Blender as well.

But I guess I should file another bug report for this, shouldn't I?
Comment by Andreas Radke (AndyRTR) - Thursday, 24 December 2009, 09:08 GMT
I'm not sure if this behavior is just wanted. You can try to find this out elsewhere. If you think it's broken file a new bug. Closing this one.

Loading...