#771227 eeepc: Black screen of death during boot

Package:
xserver-xorg-video-intel
Source:
xserver-xorg-video-intel
Description:
X.Org X server -- Intel i8xx, i9xx display driver
Submitter:
hallvor brunstad
Date:
2015-02-17 21:42:06 UTC
Severity:
important
#771227#5
Date:
2014-11-27 20:46:34 UTC
From:
To:
Dear Maintainer,

I am not quite sure what package the problem is in, but I suspect it is the
integrated Intel graphics driver.

This started happening two days ago when I rebooted my netbook.

The backlight turns off some 15-20 seconds into the boot process. This happens
on some 90% of all boots. Other times it works normally without the screen going black.

I have tried setting "nomodeset" in Grub, and it works! The only problem is poor
resolution once the computer has booted.

It would be nice if the computer could boot without the backlight going out.
#771227#10
Date:
2014-11-30 10:14:04 UTC
From:
To:
Hallvor
#771227#15
Date:
2014-11-30 10:23:11 UTC
From:
To:
control: reassign -1 xserver-xorg-video-intel

Hi Hallvor,

thanks for your bug report and follow up, reassigning accordingly...
#771227#22
Date:
2014-11-30 16:26:04 UTC
From:
To:
severity 764528 important
tag 764528 +jessie +sid +experimental
tag 771227 +jessie +sid +experimental
block 764528 763155
block 771227 763155
merge 764528 771227
thanks

All of these are probably the same bug, rendering current kernels
(jessie onward) unusable on EEEpc with older (?) Atom CPUs.

As far as I can tell, the bug may be in the kernel, not the xorg
driver (xorg shouldn't lock up the kernel; and the kernel itself
reports a BUG after all...)

IMHO, it may be reasonable to upgrade this regression to "serious"
instead of "important".

Regards,
Erich
#771227#41
Date:
2014-11-30 21:16:31 UTC
From:
To:
Please report this upstream, following instructions at
https://01.org/linuxgraphics/documentation/how-report-bugs, and let us
know the bug number for tracking.

Cheers,
Julien
#771227#48
Date:
2014-11-30 21:59:28 UTC
From:
To:
https://bugs.freedesktop.org/show_bug.cgi?id=86883

Not sure if kernel bugzilla wouldn't have been the better address,
given that the BUG_ON clearly is in kernel code. But I guess it's the
same developers anyway.

Regards,
Erich
#771227#53
Date:
2015-02-07 14:22:51 UTC
From:
To:
tag 764528 + patch jessie
tag 763155 + patch jessie
thanks

There appears to be a rather easy patch available:
http://patchwork.freedesktop.org/patch/38659/
"Quietly reject attempts to create non-pagealigned stolen objects"

Judging from the discussion, it is not a real "bug". It is an
unexpected behavior (Bios using non-aligned video buffer, supported by
the hardware, but not aligned to the driver requirements). The patch
essentially replaces the BUG_ON with a log statement and then returns
null, which is supposed to reliably work. (Without reusing the memory
the bios used.)

This probably won't be the final fix. There is some discussion here:
http://permalink.gmane.org/gmane.comp.freedesktop.xorg.drivers.intel/52613
which suggests that instead it may be desirable to reuse the bios
allocation, and just align it as desired.

But since this causes boot issues on some eeePC models (apparently, it
depends on the bios and CPU), and the patch is quite simple, it should
be considered for jessie, even if the patch isn't the long-term
solution yet. Thank you.

Regards,
Erich
#771227#58
Date:
2015-02-11 10:19:25 UTC
From:
To:
tag 764528 + patch jessie
tag 763155 + patch jessie
thanks

An improved patch has been posted, by Daniel Vetter of Intel:
https://freedesktop.org/patch/42259/

It has also been discussed on the LKML, and I hope it will be included
in the next round of pull requests of drm-next.