Bug 10965 - system hang on closing T60 panel lid when in dual screen mode 2.6.26 regression
Summary: system hang on closing T60 panel lid when in dual screen mode 2.6.26 regression
Status: CLOSED CODE_FIX
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Jesse Barnes
URL:
Keywords:
Depends on:
Blocks: 10492
  Show dependency tree
 
Reported: 2008-06-23 09:10 UTC by Franklin Piat
Modified: 2008-07-02 09:37 UTC (History)
3 users (show)

See Also:
Kernel Version: 2.6.26~rc7
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments

Description Franklin Piat 2008-06-23 09:10:24 UTC
Latest working kernel version: 2.6.25
Earliest failing kernel version: ? 2.6.26~rc5 (previous RCs not tested).
Distribution: Debian/testing
Hardware Environment: Thinkpad T60 (Intel video chipset)
Software Environment: Kernel from http://kernel-archive.buildserver.net/debian-kernel/ 

Problem Description:
 If I close my laptop's panel when I have an external monitor 
 attached and enabled, my laptop freeze... i have to power it off.


Steps to reproduce:
Plug an external screen to a Thinkpad T60 laptop (based on intel 945 chipset). Then boot the laptop, and log into gnome. Make sure that :
1. Both screen are enabled (xrandr --auto).
2. Gnome power setting has : "When laptop lid is closed" = "Blank Screen"
3. Close the laptop's panel.
=> The laptop is now frozen.

It occurs when : using linux-image-2.6.26-rc5, rc6 or rc7.
Does not occurs with : linux-image-2.6.24 , 2.6.25 from Debian/testing.

Occurs if gnome has "When laptop lid is closed" = "Blank Screen"
Doesn't occur if it's set to "Do nothing"

Occurs on my T60, with :
  Intel 945GM/PM/GMS Express Memory Controller Hub [8086:27a0]
  Intel 945GM/GMS Express Integrated Graphics Controller [8086:27a2] 
Does NOT occur on a T61, with 
  Intel GM965/GL960 Integrated Graphics Controller [8086:2a02]
  Intel GM965/GL960 Integrated Graphics Controller [8086:2a03]

Occurs when both output are enable (crt=on + lvds=on)
Doen not occurs when only one of the screens are enabled (crt=on + lvds=off OR crt=off + lvds=on).

Note :
The laptop can suspend or hibernate, then resumes correctly.

Running gnome screensaver is ok (but gnome doesn't seems to want
to put the screens in powersave mode).

More information about that bug is available at
http://bugs.debian.org/487672
Comment 1 Rafael J. Wysocki 2008-06-23 09:55:57 UTC
This entry is being used for tracking a regression from 2.6.25.  Please don't
close it until the problem is fixed in the mainline.
Comment 2 Len Brown 2008-06-23 11:17:04 UTC
Do you have
CONFIG_DRM_I915=m
and the i915 driver loaded?

Does this failure also occur with CONFIG_ACPI_VIDEO=n?
Comment 3 Jesse Barnes 2008-06-23 11:41:04 UTC
On some platforms, the BIOS will try to access disabled pipes when lid events occur.  To work around this problem, you can try using the Intel X driver's "ForceEnablePipeA" option, can you give that a try?
Comment 4 Franklin Piat 2008-06-23 16:09:26 UTC
(In reply to comment #3)
> On some platforms, the BIOS will try to access disabled pipes when lid events
> occur.  To work around this problem, you can try using the Intel X driver's
> "ForceEnablePipeA" option, can you give that a try?
> 

It works. What should I do next ?

Section "Device"
        Identifier      "Configured Video Device"
        Driver          "intel"
        Option          "ForceEnablePipeA" "true"
EndSection
Comment 5 Jesse Barnes 2008-06-23 17:23:11 UTC
Please file a bug at freedesktop.org for it.  Check out the intel man page (the ForceEnablePipeA section) for what you need to attach to the bug.

Thanks for testing.
Comment 6 Franklin Piat 2008-06-23 22:38:58 UTC
(In reply to comment #2)
> Do you have
> CONFIG_DRM_I915=m
> and the i915 driver loaded?
> 
> Does this failure also occur with CONFIG_ACPI_VIDEO=n?
> 

Yes, i915 module was loaded.

I've recompiled my (rc7) kernel with CONFIG_ACPI_VIDEO=n, but it didn't help.
Comment 7 Franklin Piat 2008-06-23 22:58:47 UTC
(In reply to comment #5)
> Please file a bug at freedesktop.org for it.  Check out the intel man page
> (the
> ForceEnablePipeA section) for what you need to attach to the bug.

Done.
https://bugs.freedesktop.org/show_bug.cgi?id=16494

Thanks for your help,

Franklin
Comment 8 Len Brown 2008-06-25 22:18:18 UTC
Thanks for the quick suggestion Jesse, and quick testing Franklin.
Clearly this one has nothing to do with ACPI, so I'm moving it
to the drivers section of bugzilla.
Comment 9 Jesse Barnes 2008-06-26 07:00:33 UTC
Franklin, since you have a good workaround now, I wonder if you could try to narrow down when this started failing.  2.6.25 included support for kernel based suspend/resume of Intel graphics devices, and so does 2.6.26, so it seems like that's not the problem.  Maybe something in the way your machine handles lid events changed?
Comment 10 Franklin Piat 2008-06-26 07:25:15 UTC
(In reply to comment #9)
> Franklin, since you have a good workaround now, I wonder if you could try to
> narrow down when this started failing.  2.6.25 included support for kernel
> based suspend/resume of Intel graphics devices, and so does 2.6.26, so it
> seems
> like that's not the problem.  Maybe something in the way your machine handles
> lid events changed?
> 

Hello,

I had started doing some git-bisecting, but it really is [CPU] time consuming... still within 11 hops, I should be down to the very one patch.

I don't know any extra test I could make, do you have any ?
Comment 11 Jesse Barnes 2008-06-26 10:15:27 UTC
Yeah, bisecting is tedious, but it's the best method I can think of for this.  Thanks!
Comment 12 Jesse Barnes 2008-07-02 09:37:17 UTC
Since this is fixed in the upstream X driver (,08903abe4dc0295c7ed7d1ff1a22e0e579540c15 of xf86-video-intel) I'm closing this out.  If your bisect finds anything, please let me know.  Thanks.

Note You need to log in before you can comment on or make changes to this bug.