Skip to contentSkip to conversation

I915 driver locks up

DISCUSSION Stephen Clark 2 participants 9 messages Dec 18, 2010 – Dec 22, 2010

lore
Share & tools
lore
#0 lore raw

Hi,

I am using 2.6.35.9-64.fc14.x86_64 running 1.3.9-1.fc14.x86_64 and am having my 945gma lock up in texture decompress according to intel_gpu_top utility, when trying to play Myst 5.

Any fixes on the horizon?

Signature · 5 lines
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty
decreases."  (Thomas Jefferson)
#1 lore raw
Quoted 8 lines · Stephen Clark

Hi,

I am using 2.6.35.9-64.fc14.x86_64 running 1.3.9-1.fc14.x86_64 and am having my 945gma lock up in texture decompress according to intel_gpu_top utility, when trying to play Myst 5.

Any fixes on the horizon?

Additional data point with 2.6.37-0.rc6.git0.1.fc15.x86_64 from rawhide I get:

[  307.256028] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[  307.258118] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -11 (awaiting 27197 at 27195, next 27199)
[  307.258975] [drm:i915_reset] *ERROR* Failed to reset chip.

and my X screen goes weird - have to switch to vty console.

Signature · 5 lines
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty
decreases."  (Thomas Jefferson)
#2 lore raw
Quoted 11 lines · Stephen Clark
Quoted 8 lines · Stephen Clark

Hi,

I am using 2.6.35.9-64.fc14.x86_64 running 1.3.9-1.fc14.x86_64 and am having my 945gma lock up in texture decompress according to intel_gpu_top utility, when trying to play Myst 5.

Any fixes on the horizon?

Additional data point with 2.6.37-0.rc6.git0.1.fc15.x86_64 from rawhide I get:

Quoted 7 lines
[  307.256028] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer 
elapsed... GPU hung
[  307.258118] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request 
returns -11 (awaiting 27197 at 27195, next 27199)
[  307.258975] [drm:i915_reset] *ERROR* Failed to reset chip.

and my X screen goes weird - have to switch to vty console.

Judging by your description, you've hit a bug in Mesa. Please verify the bug with the current stable release (of Mesa) and file a bug report, and be sure to include /sys/kernel/debug/dri/0/i915_error_state for the hang.

Thanks,
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#3 lore raw
Quoted 21 lines · Chris Wilson
Quoted 19 lines · Stephen Clark
Quoted 9 lines · Stephen Clark

Hi,

I am using 2.6.35.9-64.fc14.x86_64 running 1.3.9-1.fc14.x86_64 and am having my 945gma lock up in texture decompress according to intel_gpu_top utility, when trying to play Myst 5.

Any fixes on the horizon?

Additional data point with 2.6.37-0.rc6.git0.1.fc15.x86_64 from rawhide I get:

[  307.256028] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
elapsed... GPU hung
[  307.258118] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request
returns -11 (awaiting 27197 at 27195, next 27199)
[  307.258975] [drm:i915_reset] *ERROR* Failed to reset chip.
Quoted 10 lines
Quoted 2 lines

and my X screen goes weird - have to switch to vty console.

Judging by your description, you've hit a bug in Mesa. Please verify the bug with the current stable release (of Mesa) and file a bug report, and be sure to include /sys/kernel/debug/dri/0/i915_error_state for the hang.

Thanks,
-Chris

I had been using mesa-7.9.
I loaded mesa-7.10 similar results. There is no

/sys/kernel/debug/dri/0/i915_error_state

In fact there is nothing in the /sys/kernel/debug directory. Do I have to enable something.

With the later kernel - things are worse X becomes totally useless it looks like my lcd panel lost horizontal sync. With the 2.6.25 kernel X is still usable this makes me think at least some of the problem is kernel related.

With 2.6.37-rc6 and mesa 7.10 with hardware acceleration turned off in wine-

Dec 18 18:01:50 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.x86_64
Dec 18 18:01:52 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.i686
Dec 18 18:01:54 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.x86_64
Dec 18 18:02:02 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.x86_6
4
Dec 18 18:02:03 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.i686
Dec 18 18:02:11 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.i686
Dec 18 18:04:18 joker4 kernel: [ 9603.588113] [drm:i915_hangcheck_elapsed] *ERRO
R* Hangcheck timer elapsed... GPU hung
Dec 18 18:04:18 joker4 kernel: [ 9603.588313] [drm:i915_do_wait_request] *ERROR*
 i915_do_wait_request returns -5 (awaiting 379738 at 379735)
Signature · 5 lines
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty
decreases."  (Thomas Jefferson)
#4 lore raw
Quoted 8 lines · Stephen Clark

I had been using mesa-7.9.
I loaded mesa-7.10 similar results. There is no

/sys/kernel/debug/dri/0/i915_error_state

In fact there is nothing in the /sys/kernel/debug directory. Do I have to enable
something.

You need to "mount -tdebugfs debug /sys/kernel/debug".

Quoted 4 lines

With the later kernel - things are worse X becomes totally useless it looks like my lcd panel lost horizontal sync. With the 2.6.25 kernel X is still usable
this makes me think at least some of the problem is kernel related.

2.6.25 to present marks the transition between UMS and KMS, i.e. moving of the userspace modesetting code into the kernel. And that should not have regressed. Can you please, grab an intel_reg_dumper of the working setup and the lost-sync setup, and include a drm.debug=0xe dmesg (i.e. add drm.debug=0xe to your boot commandline)?

Quoted 9 lines

With 2.6.37-rc6 and mesa 7.10 with hardware acceleration turned off in wine-

Dec 18 18:01:50 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.x86_64
Dec 18 18:01:52 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.i686
Dec 18 18:01:54 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.x86_64
Dec 18 18:02:02 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.x86_6
4
Dec 18 18:02:03 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.i686
Dec 18 18:02:11 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.i686

Good, those drivers are recent enough that I should be able to use the error state to find the broken code.
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#5 lore raw
Quoted 36 lines · Chris Wilson
Quoted 9 lines · Stephen Clark

I had been using mesa-7.9.
I loaded mesa-7.10 similar results. There is no

/sys/kernel/debug/dri/0/i915_error_state

In fact there is nothing in the /sys/kernel/debug directory. Do I have to enable
something.

You need to "mount -tdebugfs debug /sys/kernel/debug".

Quoted 5 lines

With the later kernel - things are worse X becomes totally useless it looks like my lcd panel lost horizontal sync. With the 2.6.25 kernel X is still usable
this makes me think at least some of the problem is kernel related.

2.6.25 to present marks the transition between UMS and KMS, i.e. moving of the userspace modesetting code into the kernel. And that should not have regressed. Can you please, grab an intel_reg_dumper of the working setup and the lost-sync setup, and include a drm.debug=0xe dmesg (i.e. add drm.debug=0xe to your boot commandline)?

Quoted 10 lines

With 2.6.37-rc6 and mesa 7.10 with hardware acceleration turned off in wine-

Dec 18 18:01:50 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.x86_64
Dec 18 18:01:52 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.i686
Dec 18 18:01:54 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.x86_64
Dec 18 18:02:02 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.x86_6
4
Dec 18 18:02:03 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.i686
Dec 18 18:02:11 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.i686
Quoted 5 lines

Good, those drivers are recent enough that I should be able to use the error state to find the broken code.
-Chris

Hi Chris,

I emailed the information you requested, it was quite large, and haven't seen it show up on LKML yet. Don't know if it is being blocked by the ML or not.

Regards,
Steve

Signature · 5 lines
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty
decreases."  (Thomas Jefferson)
#6 lore raw
Quoted 35 lines · Stephen Clark
Quoted 30 lines · Chris Wilson
Quoted 8 lines · Stephen Clark

I had been using mesa-7.9.
I loaded mesa-7.10 similar results. There is no

/sys/kernel/debug/dri/0/i915_error_state

In fact there is nothing in the /sys/kernel/debug directory. Do I have to enable
something.

You need to "mount -tdebugfs debug /sys/kernel/debug".

Quoted 4 lines

With the later kernel - things are worse X becomes totally useless it looks like my lcd panel lost horizontal sync. With the 2.6.25 kernel X is still usable
this makes me think at least some of the problem is kernel related.

2.6.25 to present marks the transition between UMS and KMS, i.e. moving of the userspace modesetting code into the kernel. And that should not have regressed. Can you please, grab an intel_reg_dumper of the working setup and the lost-sync setup, and include a drm.debug=0xe dmesg (i.e. add drm.debug=0xe to your boot commandline)?

Quoted 9 lines

With 2.6.37-rc6 and mesa 7.10 with hardware acceleration turned off in wine-

Dec 18 18:01:50 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.x86_64
Dec 18 18:01:52 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.i686
Dec 18 18:01:54 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.x86_64
Dec 18 18:02:02 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.x86_6
4
Dec 18 18:02:03 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.i686
Dec 18 18:02:11 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.i686
Quoted 4 lines

Good, those drivers are recent enough that I should be able to use the error state to find the broken code.
-Chris

Quoted 8 lines

Hi Chris,

I emailed the information you requested, it was quite large, and haven't seen it show up on LKML yet. Don't know if it is being blocked by the ML or not.

Regards,
Steve

Hi Chris,

Just wondering if you received what you needed from me. I still have the info for the 2.6.37rc6 kernel that I haven't sent awaiting some kind of acknowledgment that you received the initial info I send for the 2.6.35 kernel.

Steve

Signature · 5 lines
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty
decreases."  (Thomas Jefferson)
#7 lore raw
Quoted 44 lines · Stephen Clark
Quoted 35 lines · Stephen Clark
Quoted 30 lines · Chris Wilson
Quoted 8 lines · Stephen Clark

I had been using mesa-7.9.
I loaded mesa-7.10 similar results. There is no

/sys/kernel/debug/dri/0/i915_error_state

In fact there is nothing in the /sys/kernel/debug directory. Do I have to enable
something.

You need to "mount -tdebugfs debug /sys/kernel/debug".

Quoted 4 lines

With the later kernel - things are worse X becomes totally useless it looks like my lcd panel lost horizontal sync. With the 2.6.25 kernel X is still usable
this makes me think at least some of the problem is kernel related.

2.6.25 to present marks the transition between UMS and KMS, i.e. moving of the userspace modesetting code into the kernel. And that should not have regressed. Can you please, grab an intel_reg_dumper of the working setup and the lost-sync setup, and include a drm.debug=0xe dmesg (i.e. add drm.debug=0xe to your boot commandline)?

Quoted 9 lines

With 2.6.37-rc6 and mesa 7.10 with hardware acceleration turned off in wine-

Dec 18 18:01:50 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.x86_64
Dec 18 18:01:52 joker4 yum[5244]: Updated: mesa-libGL-7.10-0.17.fc15.i686
Dec 18 18:01:54 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.x86_64
Dec 18 18:02:02 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.x86_6
4
Dec 18 18:02:03 joker4 yum[5244]: Updated: mesa-libGLU-7.10-0.17.fc15.i686
Dec 18 18:02:11 joker4 yum[5244]: Updated: mesa-dri-drivers-7.10-0.17.fc15.i686
Quoted 4 lines

Good, those drivers are recent enough that I should be able to use the error state to find the broken code.
-Chris

Quoted 8 lines

Hi Chris,

I emailed the information you requested, it was quite large, and haven't seen it show up on LKML yet. Don't know if it is being blocked by the ML or not.

Regards,
Steve

Quoted 8 lines

Hi Chris,

Just wondering if you received what you needed from me. I still have the info for the 2.6.37rc6 kernel that I haven't sent awaiting some kind of acknowledgment that you received the initial info I send for the 2.6.35 kernel.

Steve

Hi Chris,

Just wanted to say I downloaded the latest drm_fixes from Dave Airlie and tried them on 2.6.37rc7 and got the same gpu hang with the screen tearing horizontally. I have the the intel gpu dump and a cat of the i915_error_state. Together they are over 2 megabytes. Should I send them to you directly since they seem to big for the mailing list.

Regards,
Steve

Signature · 5 lines
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty
decreases."  (Thomas Jefferson)
#8 lore raw
Quoted 11 lines · Stephen Clark

Hi Chris,

Just wanted to say I downloaded the latest drm_fixes from Dave Airlie and tried them
on 2.6.37rc7 and got the same gpu hang with the screen tearing horizontally. I have
the the intel gpu dump and a cat of the i915_error_state. Together they are over
2 megabytes. Should I send them to you directly since they seem to big for the mailing
list.

Please do, or open a bug on bugs.freedesktop.org so that the error state doesn't get lost. The intel_gpu_dump should be redundant, as the same information, and more, is captured in the i915_error_state. -Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre