Bug#689268: [wheezy] Intel HD 4000 (Ivy Bridge) graphics freeze

2012-10-17 Thread Per Foreby

On 2012-10-17 05:21, Jonathan Nieder wrote:

Per Foreby wrote:


However my computer has been running without any problems for 11
days, so whatever caused this bug seems to be fixed in the 3.5.5
kernel.


Drat.  Ok.

To recap:

* Asus P8Z77-V LE.

* Newish system.  Works fine under load (e.g., Folding@Home) but
when you started normal interactive use it started to freeze a few
times a day while you were interacting with it (in particular, the
freeze happens around the same time as a keyboard or mouse action).

* The freeze is a bad one --- the fan spins down, the NIC stops
responding, caps lock doesn't light up, ctrl+alt+del and magic sysrq
have no effect.  No messages about it in netconsole.

* Happens reliably (how reliably?  80% of the time?) after a few
hours of sustained use (?)

* Logs available in the bug log.  No obvious smoking guns. ;-)

* Changing the amount of memory allocated to the integrated GPU in
BIOS doesn't change anything.

* The above describes 3.2.23-1.  Based on a week and a half of
running 3.5.5-1~experimental.1 it doesn't seem to be affected.


Correct, apart from the timing. It's been from a few hours to five days
between the freezes. But with very little interactive use during the
five days.

I can add that stressing the graphics doesn't seem to trigger the
problem.


None of the changes from 3.2 to 3.5.5 are jumping out as likely
candidates for the fix, but that's a pretty wide range.


What about the MTRR patch that Ben Hutchings mentioned above (commit
9e984bc1dffd405138ff22356188b6a1677c64c8)? Or maybe that's just a
cosmetic change? According to 
https://bugs.freedesktop.org/show_bug.cgi?id=41648, this patch was added 
in 3.5-rc1.



How reliably can you reproduce the hang on a known-bad kernel? If you
have time to try 3.2.30-1 from sid, 3.4.4-1~experimental.1 from
http://snapshot.debian.org/package/linux/ and 3.3.6-1~experimental.1
from http://snapshot.debian.org/package/linux-2.6/ then that could
help narrow down the search.


With up to five days (so far) for a freeze to occur, it might take long 
to narrow down the change, and the computer in question is semi 
production (working from home), so the freezes are very annoying. But 
I'll give it a try in the name of the good cause.


Maybe I should start with running 3.5.5 for a few weeks, just to make 
sure that the freezes really are gone?



Jonathan Nieder wrote:


 * Asus P8Z77-V LE.


This makes as good a keyword for a web search as any.

It found [1] which is not too encouraging.  Maybe memtest86+ could be
worth a try to rule some problems out.

[1] http://thread.gmane.org/gmane.linux.debian.user.french/176707/focus=176710


Memory and cpu cooling were of course the first suspects. But I'm using 
a large Arctic heatpipe cooler and have never seen higher temperatures 
than +57.0°C on any core according to lm_sensors. And memtest86+ has 
been happy. I ran an extra pass today (about 2.5 hours) just to make 
sure, and everything was blue and white.


The french discussion talks about RAM timing and voltage, but I'm not an 
overclocker (vanilla i7 3770, not 3770K) so that hardly applies.


And please note that with FAH running 24/7, the computer is always under 
heavy load, but has never frozen while running unattended. Even when 
running interactively, the freezes have never been random, but has 
happened *exactly* when I was clicking or typing something.


Btw, I found 
http://www.linuxbsdos.com/2012/10/06/ubuntu-12-04-lts-and-12-10-beta-2-on-intel-ivy-bridge-powered-computer/. 
One of the comments indicate that 3.4 fixes some sort of freeze problem 
on ivy bridge/HD4000. And the partiallysanedeveloper page that I linked 
to in the initial bug report says that the freezes are gone in 3.3.


Here are some other other discussions of freezes on similar hardware on 
debian derivatives with the same kernel generation:

http://forums.linuxmint.com/viewtopic.php?f=90t=114382
http://phoronix.com/forums/showthread.php?71895-Intel-Ivy-Bridge-On-Linux-Two-Month-Redux
http://forums.linuxmint.com/viewtopic.php?f=198t=113070
http://ubuntuforums.org/showthread.php?t=1995945

/Per


--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#689268: [wheezy] Intel HD 4000 (Ivy Bridge) graphics freeze

2012-10-17 Thread Jonathan Nieder
Per Foreby wrote:

 Correct, apart from the timing. It's been from a few hours to five days
 between the freezes. But with very little interactive use during the
 five days.

Right --- how much interactive use does it take?

I'm guessing that the time when you're not interacting the computer
doesn't affect this bug (regardless of load).

[...]
 None of the changes from 3.2 to 3.5.5 are jumping out as likely
 candidates for the fix, but that's a pretty wide range.

 What about the MTRR patch that Ben Hutchings mentioned above (commit
 9e984bc1dffd405138ff22356188b6a1677c64c8)? Or maybe that's just a
 cosmetic change?

That should be cosmetic with your card, though there's always the
possibility of something subtle happening.

[...]
 With up to five days (so far) for a freeze to occur, it might take long to
 narrow down the change, and the computer in question is semi production
 (working from home), so the freezes are very annoying. But I'll give it a
 try in the name of the good cause.

 Maybe I should start with running 3.5.5 for a few weeks, just to make sure
 that the freezes really are gone?

Selfishly, I would suggest first trying to reproduce it on a known-bad
kernel and then trying 3.3 or disabling the intel driver.

[...]
  And memtest86+ has been happy.
 I ran an extra pass today (about 2.5 hours) just to make sure, and
 everything was blue and white.

Thanks for checking.

[...]
 Btw, I found 
 http://www.linuxbsdos.com/2012/10/06/ubuntu-12-04-lts-and-12-10-beta-2-on-intel-ivy-bridge-powered-computer/.
 One of the comments indicate that 3.4 fixes some sort of freeze problem on
 ivy bridge/HD4000. And the partiallysanedeveloper page that I linked to in
 the initial bug report says that the freezes are gone in 3.3.

Yes, it would be nice to narrow this down...

If you blacklist the i915 kernel module and use the vesa X driver,
does that avoid trouble?

Thanks,
Jonathan


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#689268: [wheezy] Intel HD 4000 (Ivy Bridge) graphics freeze

2012-10-17 Thread Per Foreby

On 2012-10-18 00:02, Jonathan Nieder wrote:

Per Foreby wrote:


Correct, apart from the timing. It's been from a few hours to five days
between the freezes. But with very little interactive use during the
five days.


Right --- how much interactive use does it take?


Very little. I've had freezes so far, and all of the with almost no 
activity on the screen. Clicking a link, typing an email or closing an 
rxvt window.



That should be cosmetic with your card, though there's always the
possibility of something subtle happening.


As a rule of thumb, every bugfix in a complex enough system introduces 
another bug. So I suppose that sometimes a bugfux might accidentally fix 
another bug :)



With up to five days (so far) for a freeze to occur, it might take long to
narrow down the change, and the computer in question is semi production
(working from home), so the freezes are very annoying. But I'll give it a
try in the name of the good cause.

Maybe I should start with running 3.5.5 for a few weeks, just to make sure
that the freezes really are gone?


Selfishly, I would suggest first trying to reproduce it on a known-bad
kernel and then trying 3.3 or disabling the intel driver.


OK, I'll go back to 3.2.23 and work my way up. It might take some time 
though if the freeze doesn't behave...



If you blacklist the i915 kernel module and use the vesa X driver,
does that avoid trouble?


Does the vesa driver support 1920x1200 these days?

/Per


--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#689268: [wheezy] Intel HD 4000 (Ivy Bridge) graphics freeze

2012-10-16 Thread Per Foreby

On 2012-10-06 04:29, Per Foreby wrote:


New freeze. Last entry in the debug log was more than 10 minutes before
the freeze.

Now running 3.5-trunk-amd64 #1 SMP Debian 3.5.5-1~experimental.1 (still
with 256 MB iGPU Memory).


I was going to give it two weeks before reporting, but today we had a 
power outage so I didn't quite reach two weeks.


However my computer has been running without any problems for 11 days, 
so whatever caused this bug seems to be fixed in the 3.5.5 kernel.


/Per


--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#689268: [wheezy] Intel HD 4000 (Ivy Bridge) graphics freeze

2012-10-16 Thread Jonathan Nieder
Per Foreby wrote:

 However my computer has been running without any problems for 11 days, so
 whatever caused this bug seems to be fixed in the 3.5.5 kernel.

Drat.  Ok.

To recap:

 * Asus P8Z77-V LE.

 * Newish system.  Works fine under load (e.g., Folding@Home) but when
   you started normal interactive use it started to freeze a few times
   a day while you were interacting with it (in particular, the freeze
   happens around the same time as a keyboard or mouse action).

 * The freeze is a bad one --- the fan spins down, the NIC stops
   responding, caps lock doesn't light up, ctrl+alt+del and magic
   sysrq have no effect.  No messages about it in netconsole.

 * Happens reliably (how reliably?  80% of the time?) after a few
   hours of sustained use (?)

 * Logs available in the bug log.  No obvious smoking guns. ;-)

 * Changing the amount of memory allocated to the integrated GPU in
   BIOS doesn't change anything.

 * The above describes 3.2.23-1.  Based on a week and a half of running
   3.5.5-1~experimental.1 it doesn't seem to be affected.

None of the changes from 3.2 to 3.5.5 are jumping out as likely
candidates for the fix, but that's a pretty wide range.  How reliably
can you reproduce the hang on a known-bad kernel?  If you have time to
try 3.2.30-1 from sid, 3.4.4-1~experimental.1 from
http://snapshot.debian.org/package/linux/ and 3.3.6-1~experimental.1
from http://snapshot.debian.org/package/linux-2.6/ then that could
help narrow down the search.

Thanks again for your help and patience.

Regards,
Jonathan


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#689268: [wheezy] Intel HD 4000 (Ivy Bridge) graphics freeze

2012-10-16 Thread Jonathan Nieder
Jonathan Nieder wrote:

  * Asus P8Z77-V LE.

This makes as good a keyword for a web search as any. :)

It found [1] which is not too encouraging.  Maybe memtest86+ could be
worth a try to rule some problems out.

[1] http://thread.gmane.org/gmane.linux.debian.user.french/176707/focus=176710


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org