[i855] [karmic] X freezes with intel 855GM videocard

Sometimes it will actually freeze while GDM is loading or when GDM is actually loaded.
Turning off compiz has no effect.

Used X -configure to generate an xorg.conf file. (Changes made are made to both entries of the graphics card in the xorg.conf)
AccelMethod uxa
A freeze occurred after a while into gnome.
Gnome do was using a lot of cpu, uninstalled gnome-do.
Freeze still occurs.
DRI off --> Freeze
Option "NoAccel" "True" --> Freeze
Removed all references to two screens and cards.
AccelMethod "XAA" --> Freeze
AccelMethod "EXA" --> Freeze
AccelMethod "UXA" --> (Again because I changed the config to mono screen setup) --> Freeze
DRI "False" --> Freeze
NoAccel "True" --> Freeze
"AIGLX" "Off" --> Freeze

vesa instead of intel driver --> Doesn't even load the GDM
intel driver again
PM "off" --> Freeze
"NoMTRR" --> Freeze

Well, that is all there is on the trouble shooting page, except for the bisecting.
But I must admit, that that seems a bit well difficult and time consuming.

Hi wligtenberg,

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?

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

xorg.log as requested.
Will also post Xorg.log.old

As the status is still incomplete, could someone tell me what information I can gather to help you guys track the bug?

the same behavior here with a intel 855GM graphic chip and the karmic release. Only booting with "acpi=off" option helps so that X is not freezing anymore, the boot option "nomodeset" does not change anything, X will still freeze after a while.

So the kernel option acpi=off will actually get you a working (not freezing) system? If so, thanks for this work-around! (Will try tomorrow)

I actually tried tonight and it works!

I've also tried acpi=off and it fixes the freezing problem... But just because, at least on my system, it disables 3D acceleration completely.

I don't use desktop effects for testing, but Google Earth. If 3D is enabled, I load Google Earth and X freezes in the first minute. With acpi=off kernel parameter, GE tells me that it's using software renderer, thus no freeze, although it's painfully slow ;)

There is nothing relevant in dmesg or the X log when it freezes... Is there anything we can do to help developers track this issue?

On bug 447892 comment 23 it is described how to get potentially better batchbuffer dumps. It may also be useful to try with a recent mainline kernel from http://kernel.ubuntu.com/~kernel-ppa/mainline/ (see https://wiki.ubuntu.com/KernelTeam/MainlineBuilds) together with xorg-edgers. It would be nice if you could try that.

OK, I did try just upgrading the kernel, which didn't help.
I have tried 2.6.32-rc7 from the mainline kernels. However, the description of xorg-edgers worries me. This the laptop that I work on, so I rather have a non freezing X, without sound, 3D acceleration, than a non working laptop.
Could I just add the ppa to my sources and update the packages? And if that fails, roll them back without any harm? Or is it more difficult? (If it is this easy, and I will always be able to revert to the status I am in now, I will give it a try)

Using ppa-purge (that's found in xorg-edgers ppa), it's rather easy to rollback to the non-ppa version.

The bug applies to me too. My video card:
00:02.0 VGA compatible controller: Intel Corporation 82945G/GZ Integrated Graphics Controller (rev 02)

It needs to be corrected. A high importance is appropriate, since this bug makes the computer unusable.

I know to do an SSH to the locked machine helps, but I only have one computer at home. Can anyone help?

My computer does not have a decent 3D acceleration from Intrepid Ibex. A long year. For now, still waiting...

OK, tried with latest xorg-edgers ppa and updates installed with the 2.6.32-rc7 kernel (removed the acpi=off kernel option).
There is *some* progress. The system starts and GDM works. (as before) However when in the gnome session it freezes (as before). However, I am able to switch between tty's.
Turning the acpi=off kernel option on results in something different though. The GDM doesn't load and the screen start flickering with the terminal login.

Actually, using nomodeset and the vesa driver as explained in
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/447892
works better, than the acpi=off. My sound now works and the power button doesn't switch it off immediately.

Willem, at bug 447892 we have identified the commit to the kernel that triggered the freezes for many people on 855GM chipsets. Could you also test the before (599) and after (600) kernels to check if this is the bad commit for you as well? See comment #59 at bug 447892 for details.

[This comment goes to freeze bugs on intel 855GM]

At bug 456902 (similar freezes on 845G) there is currently an effort to find where xserver-xorg-video-intel and/or libdrm changed and so that the freezes were triggered. Brian Rogers has a PPA that he is updating libdrm and -intel until they catch a version which freezes (they first found that a standard Karmic kernel with Jaunty libdrm and -intel did not freeze). It is possible that the same is true for 855GM and this is a chance to catch a free ride from the 845G testing effort. In order to test, use the standard Karmic kernel and add Brian's PPA: https://launchpad.net/~brian-rogers/+archive/graphics-testing . See the bug 456902 comment #29 and after for more details.

Apologies if someone already reported that Jaunty with Karmic kernel also freezes. In that case, testing older -intel and libdrm in Karmic is probably pointless. I'm loosing the overview as there are many similar bug reports on 845G, 855GM and 945G/945GM, and I forget exactly what is already tested for each chipset.

I'm running into a similar problem with Karmic and the 855GM chipset. I'll try some of the suggested fixes and report more detailed debugging output later. I started having the problem upon upgrading to Karmic. When the freeze occurs, it's when an OpenGL application, like a screensaver, starts up, though only sometimes. One way to reproduce the problem is by running this from the shell:

  while [ 1 ]; do date; /usr/lib/xscreensaver/biof & sleep 5; pkill biof; sleep 1; done

Usually this will cause a freeze in ten or twenty iterations. At this point I ssh into my laptop and reboot. When I check kern.log, I see something like this shortly after the time of the freeze:

Jan 9 23:31:02 OUTOPIA kernel: [ 720.432931] INFO: task biof:2570 blocked for more than 120 seconds.
Jan 9 23:31:02 OUTOPIA kernel: [ 720.432938] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Jan 9 23:31:02 OUTOPIA kernel: [ 720.432946] biof D c08145c0 0 2570 2482 0x00000004
Jan 9 23:31:02 OUTOPIA kernel: [ 720.432958] ebb61db4 00000086 eb9ce000 c08145c0 eb9f9bb8 c08145c0 1fad12dc 0000007f
Jan 9 23:31:02 OUTOPIA kernel: [ 720.432979] c08145c0 c08145c0 eb9f9bb8 c08145c0 00000000 0000007f c08145c0 db08f000
Jan 9 23:31:02 OUTOPIA kernel: [ 720.432998] eb9f9920 f68ebc14 f68ebc18 ffffffff ebb61de0 c056fc86 c031120e f68ebc1c
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433018] Call Trace:
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433030] [<c056fc86>] __mutex_lock_slowpath+0xc6/0x130
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433043] [<c031120e>] ? get_from_free_list+0x3e/0x50
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433054] [<c056fba0>] mutex_lock+0x20/0x40
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433097] [<f8224b77>] i915_gem_gtt_pwrite_fast+0x47/0x150 [i915]
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433150] [<f81ad68b>] ? drm_gem_object_lookup+0x3b/0x50 [drm]
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433195] [<f822672a>] i915_gem_pwrite_ioctl+0x1aa/0x1d0 [i915]
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433208] [<c015f193>] ? hrtimer_forward+0x163/0x1b0
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433262] [<f81ac6c0>] drm_ioctl+0x180/0x360 [drm]
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433308] [<f8226580>] ? i915_gem_pwrite_ioctl+0x0/0x1d0 [i915]
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433321] [<c0150146>] ? run_timer_softirq+0x156/0x200
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433337] [<c013387d>] ? __enqueue_entity+0x9d/0xc0
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433349] [<c0133231>] ? __dequeue_entity+0x21/0x40
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433367] [<c01f51f3>] vfs_ioctl+0x73/0x90
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433379] [<c01f54c1>] do_vfs_ioctl+0x71/0x310
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433390] [<c01f57bf>] sys_ioctl+0x5f/0x80
Jan 9 23:31:02 OUTOPIA kernel: [ 720.433401] [<c010336c>] syscall_call+0x7/0xb

More to follow.

@17
I have today tried to 599 and 600 version.
599 works fine, 600 introduces the bug.
I will look at #18 tomorrow.

OK, I have just tried https://launchpad.net/~brian-rogers/+archive/graphics-testing

I have the following packages installed:
1:2.4.11-2+karmic3
5:2.8.0-2-gb0aa94f
Using the standard karmic kernel 2.6.31.17.30

Still produces the freeze.

Good, that is consistent with the findings of the other people who have freezes on 855GM.

Willem, I see that you tested a lot of options in xorg.conf in comment #2 and nothing stopped the freezes. Could you also try
  Option "Tiling" "false"
in the Device section? Most likely it won't make a difference, but if it does I think that is good news.

Using the option "Tiling" "false" does not make a difference it froze at login screen. (ssh worked)
Will now try to revert to the older libdrm versions and intel video drivers as instructed by Brian Rogers.

libdrm - 1:2.4.5-0ubuntu4backtrack1 combined with: xserver-xorg-video-intel - 3:2.7.0-1ubuntu1backtrack1
using stock karmic kernel: ubuntu logo shows, the screen flickers and freezes black. ssh works.
using the 599 kernel actually does the same, but no ssh!

Oops sorry, the intel driver version was: xserver-xorg-video-intel - 3:2.6.3-0ubuntu9.4backtrack1 the libdrm was:
 libdrm - 1:2.4.5-0ubuntu4backtrack1

Installed all i386 files from both packages.

Have you tried those versions along with also booting with 'nomodeset' in the kernel options? Otherwise that version of the driver probably won't do the right thing.

@Brian
No, did not try that, because I had to use nomodeset before to get this to work (with the vesa driver).
But will try again, with nomodeset tomorrow.

I retried with nomodeset kernel option
Stock kernel + xserver-xorg-video-intel - 3:2.6.3-0ubuntu9.4backtrack1
the libdrm was: libdrm - 1:2.4.5-0ubuntu4backtrack1

And it froze, though ssh was luckily available!

For anyone on Lucid, I have uploaded a version of Lucid's libdrm to my PPA which has this patch: http://cgit.freedesktop.org/mesa/drm/commit/?id=4f0f871730b76730ca58209181d16725b0c40184 .

This patch is supposed to fix many of the freeze problems that has occurred lately. The PPA is at https://launchpad.net/~gomyhr/+archive/standard .

Hi,

I just switched off quiet and splash. So my grub looks like

# If you change this file, run 'update-grub' afterwards to update
# /boot/grub/grub.cfg.

GRUB_DEFAULT=0
GRUB_HIDDEN_TIMEOUT=10
GRUB_HIDDEN_TIMEOUT_QUIET=true
GRUB_TIMEOUT="10"
GRUB_DISTRIBUTOR=`lsb_release -i -s 2> /dev/null || echo Debian`
GRUB_CMDLINE_LINUX_DEFAULT=""
GRUB_CMDLINE_LINUX=""

I have no problems any more.
greetings

oh yes, and I changed
GRUB_HIDDEN_TIMEOUT=0
to
GRUB_HIDDEN_TIMEOUT=10
bye

It seems that at least also the applications "conky" and "root-tail" made my X freeze. I used them and my X froze again, although I had splash and quiet turned off.
Now I don´t use conky and root-tail any more and everything is fine again.
greetings

@geomar:
The freezes mentioned here are related to the Intel video driver in combination with the X version.
If you experience other freezes please open a separate bug report.
The fixes you mention in #31 do not help with the freeze reported in this bug report.

There are potential fixes for this in the upstream development version. I uploaded it for Lucid to the following PPA:
https://launchpad.net/~brian-rogers/+archive/experimental

I can't simply upload this version for Karmic because it requires new versions of support libraries, but I can backport the important fixes. If someone can verify this version fixes the bug in Lucid, I'll backport the changes and we can get them in Karmic.

Also, right now this PPA contains a Karmic package that bypasses a graphical operation that was crashing i845 chips. The side effect is extremely glitchy graphics, though. It would be useful to know if the i855 problem is caused by the same operation. So testing on Karmic is also useful. If the Karmic package prevents the problem, then the problem is with i830_uxa_solid.

Hi Brian,

Just now moved to lucid. And I can confirm that this fix works.
With normal lucid, I was thrown to low graphics mode each time.
The new kernel seems to work fine.

FYI, the fix I was referring to in the earlier comment has since been included in Lucid and therefore I removed the package from my PPA. The kernel in there is a new, different fix from an upstream bug report: https://bugs.freedesktop.org/show_bug.cgi?id=27187

Glad to hear it helps.

Anyway, a newer version of the kernel patch has just been posted, so I'm sending an updated build to my PPA.

Hi Brian,

Well at 16-04-2010 the standard Lucid still crashed. Well it reverted to low graphics mode.
Using your ppa, with the new kernel makes it work fine, I don't know if that can be of any help for the Lucid release... Maybe 10.4.1?

Hi Brian,

Thanks, your kernel version 2.6.34rc3-51-generic (51.2) works on my notebook with an 855gm and a Celeron CPU (it's an old notebook :^) It complains at boot time of corrupted low memory, but the graphics are back to what I had in Karmic, 1280x768. I hope there's some way to get this into the Lucid release, or 10.04.1. Nice work, thanks again.

I can also confirm that the ppa kernel works for me in Lucid, while without it I can't boot my old laptop.

My card is:
00:02.0 VGA compatible controller [0300]: Intel Corporation 82852/855GM Integrated Graphics Device [8086:3582] (rev 02)

2.6.32-20.30 => Last one that boots, crashes with Xv
2.6.32-21.32 => It hangs when X starts
2.6.32-22.33 => It hangs when X starts
2.6.34rc3-51.2 => Works fine with KMS and Xv and everything.
One minor glitch, I'm sometimes seeing a black square over the Xv video (e.g. on totem but not on vlc), covering all of the video. In totem, when I open a menu and close it, that black square goes where the menu was. And when I switch to fullscreen it goes away so I can see the video normally. (for more details see my posts in bug #511001)

Please backport that newest patch for Lucid. :)

With 2.6.34rc3-51 my system no longer deadlocks when playing video (855GM, forced KMS makes the system boot with standard Lucid kernel, as described in bug #511001). I'll leave the computer on for a few days to see if the hangs I describe in bug #511001 are also gone.

Have any of you who have been trying 2.6.34rc3-51 experienced Xorg using abnormally much CPU time? With an idle (KDE) desktop, it seems to be consuming 10-30 times more CPU than with 2.6.32-22. It's hard to get an accurate measurement, but it's certainly qualitatively enough for my laptop's CPU fan to spin up periodically when the computer is just sitting there unused. This did not happen with 2.6.32-22 and earlier.

I can confirm that on my Asus M3N (8086:3584, 82852/855GM) Lucid works with Brian's 2.6.34rc7-51 and modesetting turned on.

CPU usage seems OK, but I noticed that video in mplayer is a bit glitchy.

On Karmic, I used a kernel from bug #447892 without modesetting.

Unfortunatelly, I have to report multpiple freezes in the last couple of weeks. If it's any help, I'll gladly send some logs the next time it happens.

@Brian Rogers: I tried the last kernel from your ppa (2.6.35rc4-v9patch1.e467e10) on another laptop with 855, and X wouldn't start unless "nomodeset" was used.

Could you please reupload 2.6.34rc3-51-generic because that works for me, but I didn't keep the .deb? Thank you!

The other methods listed in https://wiki.ubuntu.com/X/Bugs/Lucidi8xxFreezes won't give me KMS, XV and 3D simultaneously, so 2.6.34rc3-51-generic is the only way for my graphics card to work properly.

The kernels in the experimental PPA are all outdated. The new kernels are in graphics-fixes: https://launchpad.net/~brian-rogers/+archive/graphics-fixes

Try linux-image-2.6.35-ppa20+v9patch-generic. If X starts but the cursor is invisible, then try linux-image-2.6.35-ppa20+v9+cursorfix-generic.

Thank you Brian for all the help.

I tried linux-image-2.6.35-ppa20+v9patch-generic and the cursor was invisible.
I tried linux-image-2.6.35-ppa20+v9+cursorfix-generic and almost everything was working fine: KMS, 3D, XV...
glxgears even reported 308 fps instead of 260 fps that it showed with the 2.6.34rc3-51-generic kernel.

An annoying glitch with linux-image-2.6.35-ppa20+v9+cursorfix-generic though: every exactly 10 seconds, a process named kslowd000 caused a CPU spike, which resulted in the computer "hanging" for half a second - mouse not moving, screen not updating etc.

So the best solution for me was to clone my hard disk to transfer the old 2.6.34rc3-51-generic to the other laptop.
I hope that the CPU glitch above can be fixed so that I can use the newer kernel with its significant performance improvement.

Btw, why aren't those kernels mentioned in https://wiki.ubuntu.com/X/Bugs/Lucidi8xxFreezes ?
It's the only way to make those laptops work properly in Lucid...

Thanks again!

Googling revealed a thread that looks relevant to me:
http://www.gossamer-threads.com/lists/linux/kernel/1238374

One even mentions the 10 secs interval between each "hanging".
Post #15 claims to have a solution.

Thank you Brian for your work.

Actually I'm using your kernel to fix this bug on my laptop: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/568779

I'm going to dupe this against the master bug report for i855 freezing issues, and follow up on the periodic freezing issue there.