[clarkdale] entire desktop environment locks up in 9.10, i3-530 proc, built-in graphics

[KeithM]

Binary package hint: xorg

I recently built a machine, i3-530 processor, built-in graphics, Gigabyte H55M-UD2H motherboard.

I installed ubuntu 9.10 64-bit fresh on the machine. I started experiencing locks within GNOME where the entire desktop environment is frozen. Can't move the mouse, keyboard num lock doesn't work, but the machine is still running -- I can ping/ssh to it, etc.

Within a few minutes, a stack trace is dumped to /var/log/messages.

Easiest way to reproduce is just to enter Screensaver Preferences, and click on anything but the "blank screen." It happens 100% of the time.

ctrl-alt-f1, ctrl-alt-backspace: does nothing. Mouse pointer doesn't move. Keyboard LED doesn't change. I can ssh.

keith@newb:~$ lspci -nn | grep VGA
00:02.0 VGA compatible controller [0300]: Intel Corporation Clarkdale Integrated Graphics Controller [8086:0042] (rev 12)

Attached: stack trace, Xorg.0.log and Xorg.0.log.old. All in one file because I can't do multiple attachments.

Thanks for taking a look at this. Forums posts from all over indicate that a lot of people are having this issue, but perhaps with different graphics chipsets than mine.

[KeithM (keithvz)]

Regarding clarkdale integrated graphics and GPU hangs:

http://www.phoronix.com/scan.php?page=news_item&px=NzkwOA

I think this is bang-on-the-money for what's happening with me.

[Sam Stenvall (negge)]

I can confirm this bug. The same things happened today on a brand new computer with an ASUS P7H55-M PRO motherboard fitted with an Intel i3-530 CPU, both in the Live CD environment and after a successful installation (had to use the alternate installer).

After a bit of googling I found an acceptable workaround though. By chaning the last line in /etc/xdg/compiz/compiz-manager to COMPIZ_NAME="metacity" (effectively turning desktop effects off) everything seems to work just fine. I can't tell for sure if this workaround will work in the long run as I just installed the system but it's been working flawlessly for about an hour now compared to the 5-10 seconds it did before.

Could someone enlighten me/us as to whether this is a bug in the driver, in Compiz or in the kernel? For the record I'm running a stock Karmic kernel (2.6.31-19-generic).

[KeithM (keithvz)]

Sam,

Thanks for your comment.

I also found that "fix", while it did fix a lockup when ubuntu first boots up, it doesn't prevent the overall stability issues related to the video driver/or X or whatever the problem is. Depending on what you are doing (normal usage), you'll eventually trigger the lockup again. Picking a 2D or 3D screensaver seems to be a surefire way of reproducing the problem.

Note that Centos 5.3/5.4 seems to work fine on that hardware combo(i3+H55). Also, Fedora 12 Desktop worked fine. I loaded X screensaver (don't know if that is the same as on ubuntu) but also loaded a large number of screensavers, and Fedora seems to be fine. I don't know or pretend to understand all the different versions of all the utilities enough to tell why this is. I've almost exclusively used Debian, Ubuntu and derivatives (like Linux Mint) --- I'm not crazy to jump to a Redhat based release.

At the same time, this is really frustrating. I've got a brand new machine, _really_ am trying not to load Windows, and I have very little options to actually use this hardware!

Thanks

Keith

[Geir Ove Myhr (gomyhr)]

This seems like a GPU hang. There is more information about how to get the information we need (especially a batchbuffer dump) for this kind of bug at https://wiki.ubuntu.com/X/Troubleshooting/Freeze . In order to do anything about it we need to get it tested on Lucid, and since this is a very new chipset which is only officially supported from kernel 2.6.33 (Lucid ships with 2.6.32), I would like you to test this with that kernel, available from http://kernel.ubuntu.com/~kernel-ppa/mainline/?C=M;O=D (more information at https://wiki.ubuntu.com/KernelTeam/MainlineBuilds).

Sam, the bug is probably in the kernel, or -intel driver, or a combination of the two.

[madbiologist (me-again)]

Thanks to KeithM for pointing out the Phoronix article. They seem to have changed their tune slightly in paragraph 2 on page 1 of the article at http://www.phoronix.com/scan.php?page=article&item=intel_clarkdale_gpu&num=1

From my reading of several articles it does seem that the 2.6.33 kernel and the xf86-video-intel 2.10.0 driver help somewhat, although they don't completely fix things. It would be great if someone could follow Geir's request to test with the kernel in the Ubuntu PPA (as well as providing the info detailed on the troubleshooting page). I don't have a Core i3 myself, but wanted to summarise a couple of things.

[KeithM (keithvz)]

just an fyi -- i fully intend on doing extra testing to provide more info, however mother nature dumped 2+ ft of the white stuff (plus more coming) taking my power and heat with it. It might be another week until I can do this. Thanks.

[KeithM (keithvz)]

Ok, so I loaded Lucid, and then upgraded the kernel to

http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.33-rc7/linux-image-2.6.33-020633rc7-generic_2.6.33-020633rc7_amd64.deb

a couple points for the uninitiated like me:

1. Lucid seemed to require a cold reset to boot properly after installation. Warm reboots kept bringing me back to a black screen. I don't want to digress here, but this was my experience.

2. The grub configuration/version included with Lucid does not display a menu for selecting the new kernel you added with the "sudo dpkg -i *.deb" command. It, does, however, make the new kernel the new default. You can also hold down SHIFT key on bootup and you'll get a menu. This is diff from previous ubuntu versions/configs of grub.

I made an educated guess on which kernel to use, since the link Geir provided was full of directories. (note: not blaming here, I'm the ignorant one) I selected the latest RC for 2.6.33, the "v2.6.33-rc7/" directory from http://kernel.ubuntu.com/~kernel-ppa/mainline/?C=M;O=D

LONG AND SHORT:

Loading Lucid and replacing the kernel did the trick. I haven't had a single lockup since I've upgraded.

[Geir Ove Myhr (gomyhr)]

> I made an educated guess on which kernel to use, since the link Geir
> provided was full of directories. (note: not blaming here, I'm the
> ignorant one)  I selected the latest RC for 2.6.33, the "v2.6.33-rc7/"
> directory from http://kernel.ubuntu.com/~kernel-ppa/mainline/?C=M;O=D

Sorry, I usually ask for the latest -rcX , but I forgot in this case.
Good guess!

Geir Ove

[Vikram Dhillon (dhillon-v10)]

This issue was reported against karmic, so can you confirm if this issue exists with the most recent Lucid Lynx 10.04 Alpha release? ISO CD images are available at http://cdimage.ubuntu.com/releases/lucid/ . Thanks in advance.

[KeithM (keithvz)]

Vikram,

This problem definitely exists in Lucid Lynx 10.04 Alpha2, desktop amd64 bit version, that I downloaded and tested on 2/9/10.

C:\Users\xxxx\Downloads>c:\md5sum lucid-desktop-amd64.iso
64c577165a31e761a261a2d5efa53ff8 *lucid-desktop-amd64.iso

64c577165a31e761a261a2d5efa53ff8 *lucid-desktop-amd64.iso from http://cdimage.ubuntu.com/releases/lucid/alpha-2/MD5SUMS

They have file dates of 1/13/10-1/14/10.

Is there some newer version, or has something changed since 2/9/10 ??

My understanding is that the newer kernel 2.6.33 had fixed the issue, and since Lynx is shipping w/ 2.6.32 --- the problem could possibly be there (and appears to be confirmed w/ my empirical testing)

Thanks

Keith

[madbiologist (me-again)]

Sounds like it's time to get busy backporting. Three different backporting options are currently being discussed on the ubuntu-x mailing list. I hope it is not left to users to fix this by installing a LBM (the third option being discussed), although the idea at https://lists.ubuntu.com/archives/ubuntu-x/2010-February/000751.html would make this much simpler.

BTW, the newly released Lucid package xserver-xorg-video-intel 2:2.9.1-1ubuntu6 will make it easier to debug this issue with kernel 2.6.32 - according to https://lists.ubuntu.com/archives/ubuntu-x/2010-February/000762.html it triggers the apport bug info collector when X freezes. Although on second thought, that is probably of limited use, since we now know that the issue is fixed with kernel 2.6.33.

There are also a few more Ironlake bug fixes in kernel 2.6.33-rc8 - see http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.33-rc8

[madbiologist (me-again)]

I've also read a report which quoted "sources at Taiwanese motherboard manufacturers" (sounds dodgy, I know) as stating that Clarkdale is forecast to reach 20% of Intel's sales during this quarter. If this is true it's definitely something we want to support in an LTS release, preferably sooner rather than later.

[KeithM (keithvz)]

Since I upgraded to 10.04 with the 2.6.33rc7 kernel, I've seen zero problems until today. (about 12 days)

My machine locked up again with X crashing. It had crashed while one of the screensavers were running.

This sounds similar to Phoronix's experience where they said things were better, but still had occasional problems.

[KeithM (keithvz)]

FWIW, the screensaver it was on when it crashed was "GLBLUR"

[madbiologist (me-again)]

The GL in GLBLUR sounds like OpenGL. You probably already know that OpenGL is the 3D graphics acceleration system in Linux. Using this will stress the graphics stack (kernel+Mesa3D+driver) more than ordinary desktop work, and may possibly expose some remaining bugs and/or limitations. My guess is that that is what has happened here. Did apport trigger and create a bug report? If so, you can use apport-collect 516909 to attach the data from the crash to this bug report.

The final 2.6.33 kernel should be out soon. You are encouraged to try that.

[KeithM (keithvz)]

Apport did not trigger or generate a bug report as far as I can tell. Nothing related in /var/crash. Nothing in /var/log/messages.

Should I be looking elsewhere?

I'll try the final 2.6.33 whenever it is released.

Thanks

Keith

[Vikram Dhillon (dhillon-v10)]

Alright thanks for testing and reporting back, changing status to new.

[Bryce Harrington (bryce)]

Hi keithvz,

Thanks for including the attached files. Could you also include your /var/log/Xorg.0.log (or Xorg.0.log.old) from after reproducing the issue?

Please attach the output of `lspci -vvnn` too.

[This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]

[Geir Ove Myhr (gomyhr)]

You can ignore the previous message. Since you included all in one file, the script didn't find Xorg.0.log.

[Brendon Crawford (brendoncrawford)]

I would like to mention that following these instructions has completely fixed this issue for me:

http://www.linwik.com/wiki/using+the+intel+arrandale+intel+graphics+media+accelerator+hd+with+ubuntu+9.10

[KeithM (keithvz)]

Thanks for the post.

Did you have clarkdale or arrandale?

Keith

[KeithM (keithvz)]

I tried an apport-collect tonight, but it errored out (some index out of range error). Don't have time to file a bug for that too.

I've attached the .crash file. There are ~8200 of these generated in ~5 hours, like 2-3 per second.

I updated xorg according to Brendon's post and there must have been a (very) recent change, because .crash files are now being generated where they weren't a little while ago.

[KeithM (keithvz)]

Oh and note I'm on 2.6.33 final now with continuing problems.

[madbiologist (me-again)]

Sorry to hear you are still having problems with the final 2.6.33 kernel. After the significant improvement/s gained with the 2.6.33-rc's I was hoping the final kernel (in combination with a newer libdrm2) would fix the remaining problems. You must be getting frustrated.

There is a new hope. No, I'm not talking about Luke Skywalker :) Does the newly released xf86-video-intel 2.11.0 RC1 (v2.10.901) driver discussed at http://www.phoronix.com/scan.php?page=news_item&px=ODAyMQ help at all?

[KeithM (keithvz)]

madbiologist said.... "You must be getting frustrated."

Yeah, that's actually an understatement.

The bottom line is that I can't keep my hardware here idle all day, doing nothing, besides doing stability tests. I've got a brand new machine I've built which is essentially unusable. It's my fault for selecting a brand new processor, integrated graphics, etc. Bleeding edge indeed.

I'm sure this problem will eventually get addressed, but I need to start using this machine now. I've already wasted too many cycles.

CentOS 5.3 and Fedora 12 seem to work fine, and I have no idea what that tells us about the problem.

Thanks for the help & support.

Keith

[melroch (bpj)]

As an ordinary user who has been offered a desktop with a Core i3 2.93 GHz with integrated graphics (don't know which subtype but can ask), may I ask how long I will reasonably have to wait until a regular release supports that processor? TIA

[KeithM (keithvz)]

Looks like I spoke to soon. Fedora 12 is having similar problems too.

@madbiologist: Does the driver you are speaking of need compiled first, in order to be installed? xf86-video-intel 2.11.0 RC1 (v2.10.901) -- found the links, but it appears to be just source?

Briefly tried compiling, didn't work, so I gave up.

Is there some other way I can get this, or install this?

Thanks

Keith

[Geir Ove Myhr (gomyhr)]

On Wed, Mar 3, 2010 at 7:57 PM, melroch wrote:
> As an ordinary user who has been offered a desktop with a Core i3 2.93
> GHz with integrated graphics (don't know which subtype but can ask), may
> I ask how long I will reasonably have to wait until a regular release
> supports that processor? TIA

We hope that it will work well in Lucid which is due to be released on
April 29th [1]. But since we don't write the code, we depend on the
upstream code to work.

On Wed, Mar 3, 2010 at 8:47 PM, KeithM wrote:
> @madbiologist: Does the driver you are speaking of need compiled first,
> in order to be installed? xf86-video-intel 2.11.0 RC1 (v2.10.901) --
> found the links, but it appears to be just source?

The xorg-edgers PPA [2] currently has xserver-xorg-video-intel
(2:2.10.901+git20100226.a0ee9c3d-0ubuntu0sarvatt) which is identical
to 2.11.0 RC1. Of course, since this PPA will be updated continuously,
things may break in the future, but if you find the driver to work and
stop updating xorg, you may have a stable system until things are
better supported. Note that it is often problem in the kernel drivers
and not necessarily the DDX driver (which is what
xf86-video-intel/xserver-xorg-video-intel is).

[1]: https://wiki.ubuntu.com/LucidReleaseSchedule
[2]: https://launchpad.net/~xorg-edgers/+archive/ppa

[melroch (bpj)]

Re: Stable i3 support.

So the optimistic estimate is 10.4 and the conservative estimate is 10.10?
I hope I won't have to have it stand in a box/run only the preinstalled OS (for
which I have rather limited need!) for over a year!

[Geir Ove Myhr (gomyhr)]

> So the optimistic estimate is 10.4 and the conservative estimate is 10.10?

No, the realistic estimate is 10.04 and the pessimistic estimate is
10.04 updates.
This is _my_ estimate. Others may have other estimates. You won't find
any official statement in a bug report, and I don't know where/if you
can find one.

[Bryce Harrington (bryce)]

[This is an automatic notification.]

Thanks for reporting this bug to help making the Intel graphics driver
better. We hear from upstream that a number of bugs (possibly including
this one) have been fixed in the newer DRM code from the 2.6.33 kernel.
I don't know if your bug is one of the ones fixed in this release,
though, but we've prepared a PPA with this DRM update. Would you mind
installing this, rebooting, and testing if the original issue can be
reproduced with it or not?

The DRM PPA is here:

    https://edge.launchpad.net/~apw/+archive/red

Note there could be new bugs... please file these as new reports using
the command 'ubuntu-bug linux' (for kernel or DRM or KMS bugs) or
'ubuntu-bug xorg' if you suspect them to be X.org issues.

[KeithM (keithvz)]

Ok. Confirmed problem still exists in 10.04 alpha3. Locks up immediately on screensaver activation. GLBLUR seems to be the king on reproducing this.

Trying out the red PPA. No problems in ~24 hours.

[KeithM (keithvz)]

still crashing with red ppa.

This crash was different, I think X restarted. I can't be sure. All I know is that I was facing a login screen and /var/log/messages don't show a reboot.

The app-collect-error-script.py or whatever its called crashed again. Bad when the script that is supposed to report the error crashes. :)

I'm attaching crash log which automatically generated and was put into /var/crash.

[Vikram Dhillon (dhillon-v10)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 03/05/2010 06:26 PM, KeithM wrote:
> Ok. Confirmed problem still exists in 10.04 alpha3. Locks up
> immediately on screensaver activation. GLBLUR seems to be the king on
> reproducing this.
>
> Trying out the red PPA. No problems in ~24 hours.
>

Seems like Keith is still having issues in Lucid, hopefully this will be
addressed really soon in an upstream commit, if not then we'll just
write a patch here :) Changing this one to confirmed.

 status confirmed
 tag -karmic

- --
Regards,
Vikram Dhillon
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkuRzOIACgkQcoBavQdRHczDWgCgk4rLUwS4Ytpc9Ux4AOPrys/A
1aoAoInlemEbTKbJK1cd4reV4HpWWN9G
=EDiA
-----END PGP SIGNATURE-----

[Deactivated User (deactivated-user636007-deactivatedaccount)]

Did a bit of testing on the arrandale (i5-xxxM) chipsets. The red PPA seems to work after a cold boot, but not any time afterwards.

[Deactivated User (deactivated-user636007-deactivatedaccount)]

Did a bit of testing on the arrandale (i5-xxxM) chipsets. The red PPA seems to work after a cold boot, but not any time afterwards.

[Deactivated User (deactivated-user636007-deactivatedaccount)]

Hmm, wrong log file. Will repost with right log file.

[Deactivated User (deactivated-user636007-deactivatedaccount)]

Crash could not be reproduced; VT switching/X restart is still needed to get graphics working.

[KeithM (keithvz)]

Despite the initial crash, the box has been stable for about 5 days. No more entries in /var/crash. I'm still keeping an eye on it.

[KeithM (keithvz)]

Ok, definitely still crashing. Crashed again just now. Any updates or anything else to try?

Thanks

[KeithM (keithvz)]

Mar 14 20:02:22 antec kernel: [838913.849243] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
Mar 14 20:02:22 antec kernel: [838913.849250] render error detected, EIR: 0x00000000
Mar 14 20:02:22 antec kernel: [838913.849273] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -5 (awaiting 298611707 at 298611702)

Should I stay with 2.6.32 and the red PPA, or go to 2.6.33?

Thanks

[Geir Ove Myhr (gomyhr)]

KeithM, there is no reason to stay with the red PPA anymore, since the updates from there now has made it into the standard Lucid kernel 2.6.32-16. I think we have tried the different options that we have here and need to take this upstream. In order to give them the most up-to-date information to the intel developers it would be nice if you could reproduce with the newest drm-intel-next kernel from http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-next/ booted with the kernel option drm.debug=0x02 (after quiet splash), ssh in from another computer and grab /sys/kernel/debug/dri/0/i915_error_state, /var/log/Xorg.0.log, and the output of dmesg (which will have more information than usual, due to the drm.debug=0x02 option).

The automatic reporting script /usr/share/apport/apport-gpu-error-intel.py should run better now than before, but for now this manual procedure would be best.

[KeithM (keithvz)]

Thanks Geir.

I'm now on the specified kernel, running with the desired options. Just have to wait for a crash now. :)

Here's some detailed instructions:

Use this command to delete the 9238492 crash files that were generated from the last crash that you don't need:

sudo bash
find . -name '*.crash' | xargs rm

(just doing a sudo rm -rf *.crash produces "argument list too long" error)

Since grub2 is shipping w/ lucid, here's the new method for using a kernel option

You can either edit /boot/grub/grub.cfg (which is not recommended, because this will get overwritten at certain points (like installing new kernels))

and just add your drm.debug command

linux /boot/vmlinuz-2.6.33-997-generic root=UUID=7097xxxb-5a4c-4bxx-ae11-9xxxxxxxxxx52 ro drm.debug=0x02 quiet splash

OR

you can edit /etc/default/grub, add "GRUB_CMDLINE_LINUX=drm.debug=0x02" to the bottom of the file, and then "sudo update-grub"

This will append the drm debug command to all lines, but this can be removed later and updated. I'd hope that kernels that don't understand that debug option will simply ignore it. Although I'm not likely to be booting those now anyways.

Thanks

[KeithM (keithvz)]

Ok crash happened.

Attaching i915_error_state, Xorg.0.log, and output of dmesg.

[Geir Ove Myhr (gomyhr)]

Created an attachment (id=34126)
dmesg with drm.debug=0x02

Originally reported by KeithM at:
  https://bugs.launchpad.net/bugs/516909

Occationally, the GPU hangs, gets reset, then hangs again, etc. With drm.debug=0x02 on recent drm-intel-next kernel, dmesg fills up with

[34143.216468] [drm:i915_add_request], 524154
[34143.216659] [drm:i915_add_request], 524155
[34143.713488] [drm:intel_gpu_idle_timer], idle timer fired, downclocking
[34143.963261] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[34143.963268] render error detected, EIR: 0x00000000
[34143.963284] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -5 (awaiting 524155 at 512362)
[34143.963298] [drm:i915_error_work_func], generating error event
[34143.963323] [drm:i915_error_work_func], resetting chip
[34143.963533] [drm:gm45_get_vblank_counter], trying to get vblank count for disabled pipe 1

keith@newb:~$ lspci -nn | grep VGA
00:02.0 VGA compatible controller [0300]: Intel Corporation Clarkdale Integrated Graphics Controller [8086:0042] (rev 12)

[Geir Ove Myhr (gomyhr)]

Created an attachment (id=34127)
Xorg.0.log

[Geir Ove Myhr (gomyhr)]

Created an attachment (id=34128)
i915_error_state

Decoded by intel_error_decode at http://launchpadlibrarian.net/41062268/IntelErrorDecode.txt . I find two things interesting:
1. ACTHD: 0x116179c8 which is outside any of the buffers.
2. IPEHR: 0x01800002 since I have seen a lot of MI_WAIT_FOR_EVENT "Display Pipe A/B Scan Line Window Wait Enable" in this register at other bug reports on i965 and GM45 lately.

[KeithM (keithvz)]

I'm the original bug filer over at Ubuntu launchpad 516909.

I'm at your disposal to try new kernels, patches, etc. Please provide enough detail/links to howto's when replying --- I've only started messing with kernels, options, PPAs, etc recently.

Adding myself to CC: list.

Thanks

Keith

[Geir Ove Myhr (gomyhr)]

Thank you for the latest files. I have filed a bug upstream at https://bugs.freedesktop.org/show_bug.cgi?id=27115. Please register at bugs.freedesktop.org and subscribe to that bug report by adding yourself to the CC field. The intel developers may have additional questions for you there.

I attach the output of `intel_error_decode i915_error_state` that decodes the dump you uploaded to a more human readable format.

[KeithM (keithvz)]

Registered at bugs.freedesktop.org. Added to CC field.

Thanks for your help.

Keith

[Geir Ove Myhr (gomyhr)]

KeithM, I came to think about one thing. We carry a not-bleeding-edge
xserver-xorg-video-intel (2.9.1) since that gives us possibility for
UMS fallback for people who have problems on their hardware and has
been very stable. It may be possible that a newer driver will be able
to not keep resetting indefinitely. That is, it will probably still
hang, since that seems to be a kernel problem, but it may fail more
graciously. Could you try the xorg-edgers PPA which contains the
newest drivers and grab the same files as last time when/if it hangs
again? You can upload the files directly to the upstream bug report,
saying that this is from xorg-edgers.

[Chris Wilson (ickle)]

(In reply to comment #2)
> Created an attachment (id=34128) [details]
> i915_error_state
>
> Decoded by intel_error_decode at
> http://launchpadlibrarian.net/41062268/IntelErrorDecode.txt . I find two things
> interesting:
> 1. ACTHD: 0x116179c8 which is outside any of the buffers.

Looks like we are reading the wrong register for Active Head on Ironlake when grabbing error state.

> 2. IPEHR: 0x01800002 since I have seen a lot of MI_WAIT_FOR_EVENT "Display Pipe
> A/B Scan Line Window Wait Enable" in this register at other bug reports on i965
> and GM45 lately.

Similarly, I know there has been a lot of poking in this area in order to set up event triggering. Reassigning to a more knowledgeable person.

[Geir Ove Myhr (gomyhr)]

Based on a lot of similar automatic bug reports for Ubuntu, the currently active batchbuffer is probably:

0x0e043000: 0x09000000: MI_LOAD_SCAN_LINES_INCL
0x0e043004: 0x000004b0: dword 1
0x0e043008: 0x09000000: MI_LOAD_SCAN_LINES_INCL
0x0e04300c: 0x000004b0: dword 1
0x0e043010: 0x01800002: MI_WAIT_FOR_EVENT
0x0e043014: 0x54f08806: XY_SRC_COPY_BLT (rgb enabled, alpha enabled, src tile 1, dst tile 1)
0x0e043018: 0x03cc0680: format 8888, dst pitch 1664, clipping disabled
0x0e04301c: 0x00000000: dst (0,0)
0x0e043020: 0x04b00640: dst (1600,1200)
0x0e043024: 0x04781000: dst offset 0x04781000
0x0e043028: 0x00000000: src (0,0)
0x0e04302c: 0x00000680: src pitch 1664
0x0e043030: 0x0bb98000: src offset 0x0bb98000
0x0e043034: 0x02000000: MI_FLUSH
0x0e043038: 0x00000000: MI_NOOP
0x0e04303c: 0x05000000: MI_BATCH_BUFFER_END

and the correct value for ACTHD is 0x0e043014. There are many bug reports downstream with a hang in this kind of batch buffer on GM45 and 965GM, and ACTHD is always 0x.....014.
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/539804
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/539538
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/537874
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/535218
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/535010

[KeithM (keithvz)]

Yakui,

Is there any additional information you might need to fix this? Other things I can try?

Thanks
Keith

[KeithM (keithvz)]

With the 10.04 release coming this month, is there any hope in sight that this will be fixed?

I'm also watching the xorg bug site, and I don't know if a solution to the bug has even been identified yet, let alone scheduled to be fixed.

Thanks

Keith

[Richard (rd1)]

Has anyone had this problem on 9.04? Just asking because I have the same CPU and motherboard as KeithM, and found that 9.10 froze within minutes of booting live CD. However, 9.04 is fine so far, have been using it for a bit with no problems, and I can't see any mention of Jaunty on this page.

This won't help with solving the bug, but those who just need a usable Ubuntu might want to try this.

Details:

 i3-530, Gigabyte H55M-UD2H motherboard

Linux ubuntu 2.6.28-11-generic #42-Ubuntu SMP Fri Apr 17 01:57:59 UTC 2009 i686
xorg-server 2:1.6.0-0ubuntu14 (<email address hidden>)

[KeithM (keithvz)]

Richard,

I've tried just about every version, including 9.04. I've been able to reproduce in 9.04, 100% of the time.

Do me a favor. Can you go into the screensaver, pick GLBLUR as your screensaver, then just pick preview and let the screensaver run. You'll get a hang definitely within a day. It will probably be much faster.

Thanks

Keith

[madbiologist (me-again)]

The final version of the xf86-video-intel 2.11.0 driver has been released, although it may not help if you were still getting crashes when using the xorg-edgers PPA described in comment 49. xf86-video-intel 2.11.0 contains new features, performance improvements and bug fixes. See http://www.phoronix.com/scan.php?page=news_item&px=ODExMQ for more information.

[Richard (rd1)]

I'm on a dual boot system that is mostly for Windows 7 use, with Ubuntu just for emergencies, so I can't leave it over on Ubuntu 9.04. When I said it was usable - the display refresh is horribly slow, so you can see the screen repainting, but it's still possible to use it, just not so pleasant. Really hope this will be fixed in Lucid 10.04, and that there'll be an update for 9.04 and 9.10 as well.

[Geir Ove Myhr (gomyhr)]

This bug has the same type of hung batchbuffer as bug 535640 (and its many duplicates). That one was recently fixed in linux kernel version 2.6.32-19.28. KeithM, do you still get the freezes with this kernel?

[KeithM (keithvz)]

test in progress with 2.6.32-19.28.

[KeithM (keithvz)]

It may be a little early to do the Dance of Joy (http://www.youtube.com/watch?v=GfPg5LjGYz8) but everything seems to be ok.

While I've had longer series of uptime w/o a crash, nothing seems to escape the GLBLUR screensaver tests. Let me beat on it some more, and I'll post in a day or two.

Thanks

Keith

[KeithM (keithvz)]

Please see https://bugs.launchpad.net/xserver-xorg-video-intel/+bug/516909/comments/58

It seems upgrading to the 2.6.32-19.28 kernel fixes my problem.

I've run testing with it for 3-4 days, and everything is much improved.

Please change the status as appropriate.

Thanks

Keith

[KeithM (keithvz)]

Ok, I think the 2.6.32-19.28 kernel fixes the problem.

 I've run three or four days, and no crashes. Everything looks much better.

You can close out the bug as appropriate.

I'll post a comment on the upstream intel bug tracker too.

In relation to normal Ubuntu releases, is it expected that 10.04 LTS final will ship with this kernel, or some other (newer?) kernel that includes these fixes?

Thanks

Keith

[Geir Ove Myhr (gomyhr)]

(In reply to comment #7)
> It seems upgrading to the 2.6.32-19.28 kernel fixes my problem.

The patch that is included in 2.6.32-19.28 (with 2.6.33.1 drm + patches) and was intended to fix this is

  [ Jesse Barnes ]
  * SAUCE: drm/i915: don't change DRM configuration when releasing load
    detect pipe

http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-lucid.git;a=commit;h=0d2907f4bead56cff60f91068b3a3efa7149e702

I haven't seen this being applied upstream in linux-2.6.33.y or linux-2.6 git trees.

[Geir Ove Myhr (gomyhr)]

> In relation to normal Ubuntu releases, is it expected that 10.04 LTS
> final will ship with this kernel, or some other (newer?) kernel that
> includes these fixes?

There will be more fixes, but at this time only fixes that are well tested and are confirmed to fix bugs that affect Ubuntu users get in. Actually, 2.6.32-20.29 is already out. You can see the full changelog at https://launchpad.net/ubuntu/+source/linux/+changelog . The patch that fixed this problem (and many other similar problems) in -19.28 is probably
* SAUCE: drm/i915: don't change DRM configuration when releasing load
    detect pipe

[yakuizhao (yakui-zhao)]

(In reply to comment #7)
> Please see
> https://bugs.launchpad.net/xserver-xorg-video-intel/+bug/516909/comments/58
>
> It seems upgrading to the 2.6.32-19.28 kernel fixes my problem.
>
> I've run testing with it for 3-4 days, and everything is much improved.
>
> Please change the status as appropriate.
>
> Thanks
>
> Keith

thanks for the updating. So this bug will be marked as resolved.