kernel 3.2.45 Slack14 x86_64 - drm:i915_hangcheck_elapsed] *ERROR* ... GPU hung

geep999 · 06-01-2013, 05:34 AM

Hi,
Just applied the latest patches including the kernel update.
A big mistake?
PC has Intel graphics.
Typing on command line in konsole the last line became broken up - sort of "smeary". Then X went bananas and all windows locked up.
Restarting X - CTRL-ALT-BACKSPACE - I see that dmesg reports a nasty:

Code:

dmesg | grep i915
[    6.862034] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[    6.862684] i915 0000:00:02.0: setting latency timer to 64
[    6.917060] i915 0000:00:02.0: irq 53 for MSI/MSI-X
[    7.651517] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0
[ 6467.574708] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[ 6467.574717] [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
[ 6467.578198] [drm:i915_wait_request] *ERROR* i915_wait_request returns -11 (awaiting 1668973 at 1668972, next 1668974)

If it does this a few more time I'll be reverting to the original kernel.

Cheers,
Peter

willysr · 06-01-2013, 07:50 AM

Did you use build 1 or build 2? Pat has rebuilt the kernel by removing the commit that causes this problem

digger95 · 06-01-2013, 07:58 AM

The second build is still causing issues for some Intel users and Pat is considering another rebuild. See the last few posts in this thread:

http://www.linuxquestions.org/questi...ml#post4962341

Philip Lacroix · 06-01-2013, 10:25 AM

Quote:

Originally Posted by geep999

Just applied the latest patches including the kernel update. PC has Intel graphics. Typing on command line in konsole the last line became broken up - sort of "smeary". Then X went bananas and all windows locked up.

Some of us could easily fix the problem by removing commit 393143615d9f2f581d87387268dc11b95adc339c from the 3.2.45 kernel source, as suggested by Patrick in the thread mentioned by digger95: you might want to read and post there. This issue is not fixed by the latest Linux 3.2.46 update. There are some useful suggestions in this thread as well.

Best regards,
Philip

geep999 · 06-02-2013, 06:15 AM

Thanks for all the replies. At least I'm not alone.
I got the problem after installing this patch (and all the ones before it):
Wed May 22 14:11:13 UTC 2013
patches/packages/linux-3.2.45/*: Rebuilt.

As long the problem remains very intermittent I won't bother to revert to older kernel.
And I'll keep on eye on the links listed above, and the Slackware changelog, for a final fix.

Cheers,
Peter

ruario · 06-02-2013, 06:17 AM

The rebuild from May 22nd was to fix another issue. You will still need to remove the commit that Philip mentioned before your issue will go way.

TommyC7 · 06-02-2013, 09:15 AM

I think this might be fixed in 3.2.46 (I haven't checked the source code, though).

I'm currently running 3.2.46 also using the i915 driver and this is the output of my `dmesg | grep i915':

Code:

i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
i915 0000:00:02.0: setting latency timer to 64
i915 0000:00:02.0: irq 46 for MSI/MSI-X
[drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0

ruario · 06-02-2013, 09:24 AM

Quote:

Originally Posted by TommyC7

I think this might be fixed in 3.2.46 (I haven't checked the source code, though).

It isn't. Quotes from above and the main thread.

Quote:

Originally Posted by Philip Lacroix

This issue is not fixed by the latest Linux 3.2.46 update.

Quote:

Originally Posted by Philip Lacroix

Ok, Linux 3.2.46 is now available on kernel.org, however the patch doesn't revert the lines discussed above. I decided to try it anyway, so I updated the original 3.2.45 sources with today's 3.2.46 incremental patch, rebuilt the kernel, modules, initrd and so on, and the previous issues are here again. Just FYI

Quote:

Originally Posted by ruario

the 3.2.46 ChangeLog makes no reference to either of the troublesome commits (e12a2d53ae45a69aea499b64f75e7222cca0f12f and 393143615d9f2f581d87387268dc11b95adc339c).

Philip Lacroix · 06-02-2013, 10:39 AM

Quote:

Originally Posted by ruario

The rebuild from May 22nd was to fix another issue. You will still need to remove the commit that Philip mentioned before your issue will go way.

Yeah, and that's the only commit I removed from the original 3.2.45 kernel: I didn't even try the second build, as I was doing very well with 3.4.47 and 3.8.13. Now I'm back to 3.2.45, as removing that commit made the kernel rock-stable again. I really like the way Slackware lets you try out things, with no dependency management tools fighting against you.

@geep999, you might want to rebuild the kernel after having that commit removed. You can even delete those few lines manually: look at drivers/gpu/drm/i915/intel_ringbuffer.c in your kernel source, lines 411-414 (for both 3.2.45 and 3.2.46, as the file wasn't touched by the last update). You can gather useful information in the mentioned threads, and if you don't know how to build a kernel Ruarí's post here can be a good start, along with this document from the Slackware Documentation Project.

Philip

ruario · 06-04-2013, 05:18 AM

New kernel package provided by Pat with the second problematic commit removed.

Quote:

Originally Posted by Philip Lacroix

Now I'm back to 3.2.45, as removing that commit made the kernel rock-stable again.

I also decided to go back to 3.2.45 (with the two problematic commits removed), as I like to stay as close to vanilla Slackware as possible. It helps when helping with others problems or when they are helping with mine if you have very similar setups.

Philip Lacroix · 06-04-2013, 07:37 AM

Quote:

Originally Posted by ruario

I like to stay as close to vanilla Slackware as possible. It helps when helping with others problems or when they are helping with mine if you have very similar setups.

That's true. On the other hand, in a situation like this it was also very helpful to know that alternate solutions were viable. I didn't want to downgrade to 3.2.29, and both 3.4.47 and 3.8.13 kernels proved to be excellent, at least here.

By the way, I had done a custom build with the second commit reverted and the i915 module built in, along with its dependencies. This way no updated modules were required if the older kernel-modules-3.2.45 package was there. I just replaced the original kernel image with my build on a second Intel box with the crashing 3.2.45 upgrade, I run lilo and the issues got fixed. I was managing to upload this somewhere, then the official rebuild was available and made it useless.