Bug#689268: linux-image-3.2.0-3-amd64: Intel HD 4000 (Ivy Bridge) graphics freeze

[Ingo <ingo.steiner@gmx.net>]

I can confirm this bug here too and found a temporarly workaround.

Ivy-Bridge i5-3570K on Intel DH77EB MoBo (H77 chipset), latest BIOS
EB0089.BIO.

These freezes happend most of the time when hitting a link in Iceweasel.
They are so severe that even the MoBo reset button does not respond
immediately, I had to hit it several times. Additionally when PC stalls,
monitor continues to display the frozen desktop (via displayport) and
power consumption rises from idle 38 watts to constant 86 watts - verry
dangerous if also fan regulation fails. Upon next boot I get "orphaned
inodes" - very much the same as Per Foreby describes.

System here is Wheezy-amd64 with
- kernel 3.2.23-1
- xserver-xorg-video-intel 2:2.19.0-5

Now I'll give the history what I did try with which result. Sorry if
it's not scientific, but probably helps to pin down the root cause.


It started when I was examining my logs and discovered the messages:

  [drm] MTRR allocation failed.  Graphics performance may suffer.

Checking mtrr's showed:

cat /proc/mtrr
reg00: base=0x000000000 (    0MB), size= 8192MB, count=1: write-back
reg01: base=0x200000000 ( 8192MB), size=  512MB, count=1: write-back
reg02: base=0x0e0000000 ( 3584MB), size=  512MB, count=1: uncachable
reg03: base=0x0dc000000 ( 3520MB), size=   64MB, count=1: uncachable
reg04: base=0x0db800000 ( 3512MB), size=    8MB, count=1: uncachable
reg05: base=0x21f800000 ( 8696MB), size=    8MB, count=1: uncachable
reg06: base=0x21f600000 ( 8694MB), size=    2MB, count=1: uncachable

So I decided to boot with kernel parameter enable_mtrr_cleanup which
seemed to solve the problem:

reg00: base=0x000000000 (    0MB), size= 2048MB, count=1: write-back
reg01: base=0x080000000 ( 2048MB), size= 1024MB, count=1: write-back
reg02: base=0x0c0000000 ( 3072MB), size=  512MB, count=1: write-back
reg03: base=0x0db800000 ( 3512MB), size=    8MB, count=1: uncachable
reg04: base=0x0dc000000 ( 3520MB), size=   64MB, count=1: uncachable
reg05: base=0x100000000 ( 4096MB), size= 4096MB, count=1: write-back
reg06: base=0x200000000 ( 8192MB), size=  512MB, count=1: write-back
reg07: base=0x21f600000 ( 8694MB), size=    2MB, count=1: uncachable
reg08: base=0x21f800000 ( 8696MB), size=    8MB, count=1: uncachable
reg09: base=0x0e0000000 ( 3584MB), size=  256MB, count=1: write-combining

At least since then (don't know whether before as well) the
freezes/crashes were observed, sometimes more than once a day.

Did try a lot to find the root cause and finally found following workaround:

Default setting in the BIOS of the DH77EB for video agp-aperture is
"max" (values of 64, 128, 256 and 512MB are offered as options).
I played around with different BIOS settings and observed that these
settings are not respected by the i915 module. Dmesg always reports
256MB for the aperture:

dmesg | grep agp
Linux agpgart interface v0.103
agpgart-intel 0000:00:00.0: Intel Ivybridge Chipset
agpgart-intel 0000:00:00.0: detected gtt size: 2097152K total, 262144K
mappable
agpgart-intel 0000:00:00.0: detected 65536K stolen memory
agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xe0000000

So I decided to set the BIOS AGP-aperture to 256MB as well and removed
the kernel parameter 'enable_mtrr_cleanup'.

I now get for the mtrr's:

reg00: base=0x000000000 (    0MB), size= 8192MB, count=1: write-back
reg01: base=0x200000000 ( 8192MB), size=  512MB, count=1: write-back
reg02: base=0x0e0000000 ( 3584MB), size=  512MB, count=1: uncachable
reg03: base=0x0dc000000 ( 3520MB), size=   64MB, count=1: uncachable
reg04: base=0x0db800000 ( 3512MB), size=    8MB, count=1: uncachable
reg05: base=0x21f800000 ( 8696MB), size=    8MB, count=1: uncachable
reg06: base=0x21f600000 ( 8694MB), size=    2MB, count=1: uncachable

still suffering graphics performance and "mtrr missmatch" according to
dmesg:

mtrr: type mismatch for e0000000,10000000 old: write-back new:
write-combining
[drm] MTRR allocation failed.  Graphics performance may suffer.

But since then I have never obseved any cras/freeze for days now.

If more information is required, please let me know (logs did never
contain any information related to the crashes). My suspicion for the
cause is some memory or communication missmatch between BIOS-setting,
mtrr's and agp-aperture.