Bug 14844 - Hard system lock with textured video
Summary: Hard system lock with textured video
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/Radeon (show other bugs)
Version: 7.3 (2007.09)
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: xf86-video-ati maintainers
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-03-05 14:55 UTC by ezerotven+fdo
Modified: 2008-04-18 08:27 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
xorg.conf (3.81 KB, application/octet-stream)
2008-03-05 14:56 UTC, ezerotven+fdo
no flags Details
xorg log (11.57 KB, application/x-bzip)
2008-03-06 01:42 UTC, ezerotven+fdo
no flags Details

Description ezerotven+fdo 2008-03-05 14:55:13 UTC
Hi,

I'm using xorg from debian unstable with the xserver-xorg-video-ati driver from experimental (version: 6.8.1~git20080302.a4398ac3-1). I have a hp nw8440 laptop with Radeon Mobility X1600.

Sometimes when I use mplayer with the Xv target to watch some video my system locks hard. Alt+sysreq does not work. Unfortunately I have not found a deterministic way to reproduce the problem, yet.

Any idea?

Thanks, Laszlo
Comment 1 ezerotven+fdo 2008-03-05 14:56:07 UTC
Created attachment 14870 [details]
xorg.conf
Comment 2 Alex Deucher 2008-03-05 16:09:31 UTC
Textured video on r5xx and rs6xx is still experimental.  Are you using the drm?  Can you post your xorg log?
Comment 3 ezerotven+fdo 2008-03-06 01:42:23 UTC
Created attachment 14874 [details]
xorg log

I do not use the drm. I know textured video is experimental, but I'm willing to do testing to finally be able to get rid of fglrx.
Comment 4 Alexander Sabourenkov 2008-03-27 12:41:51 UTC
I also experience this bug on an RV530, AGP. 

Today I pulled mesa/drm and driver and got this:

With stock drm (ubuntu gutsy 2.6.22) - two lockups on login, one time I managed to open an xterm before lockup.

With git drm (kernel modules only, I have not installed libdrm) - no signal on monitor, Xorg process eats all cpu, unkillable. 


Comment 5 Alex Deucher 2008-03-27 13:02:16 UTC
Please try with the drm.  The MMIO path does not seem to be completely stable at the moment on all chips.  Also you might try enabling EXA as well:
Option "AccelMethod" "EXA"

Unfortunately, AGP cards will need proper support for the AGP->PCIE bridge before they will work properly with the drm.
Comment 6 Alexander Sabourenkov 2008-03-27 14:04:13 UTC
Device section of the xorg.conf in question:

Section "Device" 
        Identifier      "X1600Pro"
        Driver          "radeon"
        Option "AccelMethod" "EXA"
        Option "MigrationHeuristic" "greedy" 
EndSection

As I wrote in the previous comment, git head with drm either spins or locks up. 
I understand that this is an incomplete report, will post something useful next time. I'll try git-head also on an PCIe RV570 later this week for comparison.

I have several questions regarding testing so that its results will be most useful:

Can AGPMode option and agp-related BIOS settings help in rv530 case?
Is libdrm recompile needed? 
What binaries depend on libdrm? Should they be recompiled too?
Does X server version matter?
Does (linux) kernel version matter?

Comment 7 Alex Deucher 2008-03-27 14:11:57 UTC
(In reply to comment #6)
> Device section of the xorg.conf in question:
> 
> Section "Device" 
>         Identifier      "X1600Pro"
>         Driver          "radeon"
>         Option "AccelMethod" "EXA"
>         Option "MigrationHeuristic" "greedy" 

You shouldn't need the migration heuristic option now that the driver has EXA composite support.

> EndSection
> 
> As I wrote in the previous comment, git head with drm either spins or locks up. 
> I understand that this is an incomplete report, will post something useful next
> time. I'll try git-head also on an PCIe RV570 later this week for comparison.
> 
> I have several questions regarding testing so that its results will be most
> useful:
> 
> Can AGPMode option and agp-related BIOS settings help in rv530 case?

The bios options may be useful eventually, but the problem is the chip on the card that bridges AGP to PCIE has to be set up properly to translate between the two interfaces.

> Is libdrm recompile needed? 

You'll new to rebuilt that in order to compile the latest kernel modules.

> What binaries depend on libdrm? Should they be recompiled too?

You don't have to worry about them.

> Does X server version matter?

In this case, no.

> Does (linux) kernel version matter?

Only in that you need the headers from your current kernel in order to build the latest drm from git.  Depending on the age of your kernel and what your distro has included/changed, you may have to make some minor changes to get things to build properly.
Comment 8 ezerotven+fdo 2008-04-18 02:23:16 UTC
Hi,

I've finally managed to test this using drm from git head, and now textured video works for me.

Thanks,Laszlo


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.