[i855] Karmic / i855GM / Fujistu Siemens Amilo M7400 : Freeze shortly after X startup on Intel i855

I can confim this behaviour on Maxdata Pro i600 Notebook. After upgrading from 9.04 to 9.10 the display freezes and cannot be restarted at all. The system is fully available from network (i.e. ssh shell) and can be shut down or restarted. System load is ~1 although 100%idle on CPU.

There is a stack trace in the messages log which shows, that this must be a problem of the i915 driver:

Nov 8 12:17:33 seal kernel: [ 840.420086] i915/0 D c08145c0 0 285 2 0x00000000
Nov 8 12:17:33 seal kernel: [ 840.420101] f6b33f04 00000046 f6868000 c08145c0 f6b71bb8 c08145c0 302255c5 00000092
Nov 8 12:17:33 seal kernel: [ 840.420121] c08145c0 c08145c0 f6b71bb8 c08145c0 30224573 00000092 c08145c0 f6bec000
Nov 8 12:17:33 seal kernel: [ 840.420141] f6b71920 f68f1414 f68f1418 ffffffff f6b33f30 c056f776 f7070c90 f68f141c
Nov 8 12:17:33 seal kernel: [ 840.420160] Call Trace:
Nov 8 12:17:33 seal kernel: [ 840.420182] [<c056f776>] __mutex_lock_slowpath+0xc6/0x130
Nov 8 12:17:33 seal kernel: [ 840.420193] [<c056f690>] mutex_lock+0x20/0x40
Nov 8 12:17:33 seal kernel: [ 840.420256] [<f81abc0a>] i915_gem_retire_work_handler+0x2a/0x70 [i915]
Nov 8 12:17:33 seal kernel: [ 840.420273] [<c0157a7e>] run_workqueue+0x6e/0x140
Nov 8 12:17:33 seal kernel: [ 840.420318] [<f81abbe0>] ? i915_gem_retire_work_handler+0x0/0x70 [i915]
Nov 8 12:17:33 seal kernel: [ 840.420331] [<c0157bd8>] worker_thread+0x88/0xe0
Nov 8 12:17:33 seal kernel: [ 840.420344] [<c015c280>] ? autoremove_wake_function+0x0/0x40
Nov 8 12:17:33 seal kernel: [ 840.420357] [<c0157b50>] ? worker_thread+0x0/0xe0
Nov 8 12:17:33 seal kernel: [ 840.420368] [<c015bf8c>] kthread+0x7c/0x90
Nov 8 12:17:33 seal kernel: [ 840.420378] [<c015bf10>] ? kthread+0x0/0x90
Nov 8 12:17:33 seal kernel: [ 840.420391] [<c0104007>] kernel_thread_helper+0x7/0x10

Please see the full dmesg attached.

00:02.0 VGA compatible controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)
        Subsystem: ASUSTeK Computer Inc. Device 1742
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
        Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx+
        Latency: 0
        Interrupt: pin A routed to IRQ 16
        Region 0: Memory at f0000000 (32-bit, prefetchable) [size=128M]
        Region 1: Memory at feb00000 (32-bit, non-prefetchable) [size=512K]
        Region 2: I/O ports at dc00 [size=8]
        Capabilities: <access denied>
        Kernel driver in use: i915
        Kernel modules: i915

Thanks for debugging this issue. I hope my information helps.
 - Dan

I just realised, this also happens on Karmic Koala Live-CD (just wanted to do a fresh install). So the i915/855 problem is not only depending on an upgrade from Intrepid.

HTH
 - Dan

Pierre, I have forwarded this bug report upstream so that the intel developers may have a look at the batchbuffer dump. Please regester at freedesktop.org and add yourself to the CC field of the upstream report, so that you may answer any questions that they might have.

https://bugs.freedesktop.org/show_bug.cgi?id=25086

Pierre, on some other ubuntu freeze bugs on 855GM it is reported that booting with the kernel parameter acpi=off works around the problem (Esc at boot to show the Grub menu, 'e' to edit the default entry, add acpi=off after quiet splash, Ctrl+C to boot IIRC). Could you check if this is the case for you also?

Another tests I would like you to do:
- Check whether it occurs without KMS, that is, instead of acpi=off, use the option nomodeset

There are a couple of more options that you can try in xorg.conf at https://wiki.ubuntu.com/X/Troubleshooting/Freeze#Narrow%20Subsystem%20it%20Occurs%20in . If any of them stop the freezes or reduce the frequency of them that is also a useful data point, since it may tell us which part of the driver is responsible.

Finally, there's a chance that a newer kernel might have some fixes that resolves this. You may get newer mainline builds at http://kernel.ubuntu.com/~kernel-ppa/mainline/ and there is some more information about those builds at https://wiki.ubuntu.com/KernelTeam/MainlineBuilds .

Even if none of the mentioned tests changes anything, that is a very useful thing to know.

Kernel option 'nomodset' does not solve the problem, neither does an modset=0 in newly created /etc/modules.d/i915-kms.conf. Driver crash is not due to enabled KMS im my opinion.

Dan, we (and more importantly, upstream) have a policy of only one
reporter per bug report. This is because xorg-bugs that look exactly the
same can have different causes. It soon gets confusing when we have
files from different reporters in the same bug report, especially if it
turns out that the logs are incompatible.

We appreciate your information, but we would prefer it in a separate bug
report. If the issues turn out to be the same, it can be marked as a
duplicate, which means that we have two linked bug reports of the same
issue, but it is easy to keep track of which files come from where.

There are already quite a few bug reports on 855GM freezes
(see https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bugs?field.tag=855gm%20freeze&field.tags_combinator=ALL), but since you are providing high quality input (in a place where it is not very useful), it would be nice if we could have one from you as well. Please take a look at the other 855GM freeze bugs and take a look at the different tests that are useful and include that in your report (and the comments you made here). Make sure to make it a complete report with /var/log/Xorg.0.log, full dmesg output and output of `lspci -vvnn` as separate attachments. If you get all the information in there right away, we avoid a few rounds of back-and-forth communication and can put your information to use right away.

Sorry for the time to answer

See above the result of my last tests

-2.4 intel driver from the x-retro repository (xserver-xorg-video-intel-2.4_2.4.1-1ubuntu11~ppa1_i386.deb) => OK works, good perf
-acpi=off boot option => OK works but poors performances (i don't know why)
-nomodeset boot option => KO freezes quickly after X startup
-"options i915 modeset=0" in /etc/modprobe.d/i915-kms.conf => KO freezes quickly after X startup (normal i think it's the same conf as previous)
-option "DRI" "Off" in xorg.conf => KO freezes quickly after X startup
-Option "PM" "Off" in xorg.conf => KO freezes quickly after X startup
-Option "NoMTRR" in xorg.conf => KO freezes quickly after X startup

i don't try because ignoreby drivers version > 2.8 :
Option "AccelMethod" "xxx" - Try "XAA", "EXA" (ignored on -intel > 2.8.0)
Option "Accel" "Off" - turns off the 2D acceleration (ignored on -intel except for i810 and i815 chipsets)

Thanks

I have a theory as for why acpi=off works, but not nomodeset or "DRI"
"off". As far as I remember, the xorg.conf option "DRI" "off" is
ignored when KMS is used (cf. comment # 6, DRI is responsible for KMS,
so it does not make sense to turn it off). So what acpi=off does is
probably to disable DRI, and therefore also disable KMS. If this
theory is right, using the kernel option nomodeset AND having Option
"DRI" "off" in xorg.conf will stop the freezes.

Could you check if this is the case? To understand better what is
happening with acpi=off, it would be nice to have /var/log/Xorg.0.log
for both the case acpi=off (without xorg.conf) and the case nomodeset
with Option "DRI" "off" in xorg.conf.

> i don't try because ignoreby drivers version > 2.8 :
> Option "AccelMethod" "xxx" - Try "XAA", "EXA" (ignored on -intel > 2.8.0)
> Option "Accel" "Off" - turns off the 2D acceleration (ignored on -intel except for i810 and i815 chipsets)

Yes, I added those comments recently so that people with intel
wouldn't try those options to no avail. I guess I should also add that
the DRI option is ignored when KMS is used once we have verified that.

Actually, I'm not quite sure when the "DRI" option is taken into account and when it's not. On my computer (965GM) it is not ignored, Xorg.0.log says
(**) intel(0): Option "DRI" "off"
but glxinfo says
direct rendering: Yes
and yet my OpenGL screensaver (Busy Spheres) is down to about 1 fps as opposed to something reasonable without the option, so something must have been turned off. The comment # 6 I referred to in the previous comment was in bug 447892 and not this one. In that case the DRI option is clearly ignored. Maybe it depends on the chipset. It would therefore be nice to have your Xorg.0.log with KMS (i.e. no kernel option) and Option "DRI" "off" in xorg.conf in addition to the ones mentioned in the previous comment.

Hi,

see above the traces files and comments according to the last tests asked :

No options -> Xorg.0.log-nothing -> KO freezes
acpi=off -> Xorg.0.log-acpi-off -> OK but poor perf
dri=off in xorg.conf and nomodeset option set in grub parameter -> line Xorg.0.log-nomodeset_grub-dri_off -> KO freezes
dri=off in xorg.conf and modeset=0 in modprobe.d -> Xorg.0.log-nomodeset_modprobe-dri_off -> KO freezes

The four Xorg.0.old log files are in the tgz archive.

Thanks

OK i missed the lastest test :

only DRI=OFF in xorg.conf -> Xorg.0.log-only-dri_off -> KO freezes

Pierre, 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.

Hi Geir ,

i cannot access the address http://www.kvante.info/855GMfreeze/ to download the two deb packages.

I have a

Forbidden

You don't have permission to access /855GMfreeze/ on this server.

message.

Could you give me new address.

Thank you

> i cannot access the address http://www.kvante.info/855GMfreeze/ to
> download the two deb packages.

Sorry about that! I was supposed to copy the .htaccess file to another
directory (for another bug) and ended up moving it instead. It's fixed
now.

Geir Ove

Good,

i used the nopae kernel version

599 -> No freeze (too good!!, to used my computer normaly with acceleration)
600 -> KO freeze

Hope the patch will be quickly available

Thank you

[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.