Nouveau - Jun 2012 - [Bug 50571] New: nouveau crashes with GeForce GT 520

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #1 from Gedalya <gedalya at gedalya.net> 2012-06-01 00:11:45
UTC ---
Created attachment 62358
  --> https://bugs.freedesktop.org/attachment.cgi?id=62358
Kernel messages - partedmagic - starting Xorg

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

             Bug #: 50571
           Summary: nouveau crashes with GeForce GT 520
    Classification: Unclassified
           Product: xorg
           Version: unspecified
          Platform: x86-64 (AMD64)
        OS/Version: Linux (All)
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Driver/nouveau
        AssignedTo: nouveau at lists.freedesktop.org
        ReportedBy: gedalya at gedalya.net
         QAContact: xorg-team at lists.x.org


Created attachment 62357
  --> https://bugs.freedesktop.org/attachment.cgi?id=62357
Kernel messages - starting gdm3

Original report: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=675302

The system hangs as soon as starting xorg when using the nouveau module. The
non-free nvidia driver or just the fbdrv work fine.

I'm including logs for what happens when starting gdm3 under debian using a
3.3.6 kernel and when starting Xorg under partedmagic also using linux 3.3.6

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

Gedalya <gedalya at gedalya.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |http://bugs.debian.org/6753
                   |                            |02

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

Jonathan Nieder <jrnieder at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #62357|application/octet-stream    |text/plain
          mime type|                            |

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

Jonathan Nieder <jrnieder at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #62358|application/octet-stream    |text/plain
          mime type|                            |

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #2 from Jonathan Nieder <jrnieder at gmail.com> 2012-06-01
00:27:50 PDT ---
More details from http://bugs.debian.org/675302:

1. Starting and using X with the fbdev driver works fine.

2. "startx /usr/bin/xterm" to start a minimal X session using the
nouveau
driver freezes, too.

| Now tried running startx /usr/bin/xterm with nouveau,
| 
| [   82.427553] [drm] nouveau 0000:01:00.0: PDISP: DCB for 6/0xbad00103 not
found
| [   82.428536] [drm] nouveau 0000:01:00.0: PDISP: DCB for 0/0xbad00103 not
found
| [   82.429483] [drm] nouveau 0000:01:00.0: Table 0x0103 not found for 0/2,
using first
|
| I kept a previously opened ssh connection.
| When starting X, the screen went black, but didn't totally lock up until
| I killed the X process from SSH. No further netconsole output, the
| machine went totally dead.

3. This is a regression.

| A while back (like around linux 3.0.0) nouveau and gnome 2.30 did work 
| on this same machine

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #3 from Marcin Slusarz <marcin.slusarz at gmail.com>
2012-06-01 12:42:40 PDT ---
Kernel crashes on this line in nvd0_display.c/evo_wait:
disp->evo[id].ptr[put] = 0x20000000;
because "put" has unexpectedly large value: 2eb40040 / 2eb7c400 /
2eb40040.
I bet evo id is wrong and we read wrong register.

I'm not sure how to proceed from here. For now please test 3.4 kernel,
attach
new dmesg and vbios [1] (last messages indicate it might be related to vbios
parsing).

[1] http://nouveau.freedesktop.org/wiki/DumpingVideoBios

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #4 from Jonathan Nieder <jrnieder at gmail.com> 2012-06-01
12:55:45 PDT ---
Instructions for testing a more recent kernel:

# prerequisites
apt-get install git build-essential

# get a copy of the kernel history, if you do not already have it
git clone \
  git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

# get latest rev from nouveau tree
git remote add nouveau \
  git://anongit.freedesktop.org/git/nouveau/linux-2.6
git fetch nouveau
git checkout nouveau/master

# configure
cp /boot/config-$(uname -r) .config; # current configuration
scripts/config --disable DEBUG_INFO

# optional: minimize configuration
# This works by disabling all modules that are not loaded,
# which are generally drivers for hardware you don't have.
# Make sure nouveau is loaded before doing this.
make localmodconfig

# build, test
make deb-pkg; # optionally with -j<num> for parallel build
dpkg -i ../<name of package>; # as root
reboot

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #9 from Gedalya <gedalya at gedalya.net> 2012-06-13 17:16:18
UTC ---
Created attachment 62999
  --> https://bugs.freedesktop.org/attachment.cgi?id=62999
Linux 3.4.1 - works

Yea, that works too.

OK, anything else I can do to help?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #5 from Gedalya <gedalya at gedalya.net> 2012-06-13 10:36:23
PDT ---
Created attachment 62981
  --> https://bugs.freedesktop.org/attachment.cgi?id=62981
ROM of my "ASUS ENGT520 Silent/DI/1GD3(LP) GeForce GT 520"

Sorry for such a long delay.
This was dumped using vbtracetool.
NVIDIA non-free drivers were in use at the time, hope it doesn't matter.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #6 from Gedalya <gedalya at gedalya.net> 2012-06-13 15:42:15
PDT ---
On 6/1/2012 3:55 PM, bugzilla-daemon at freedesktop.org
wrote:
> # get latest rev from nouveau tree
> git remote add nouveau \
>    git://anongit.freedesktop.org/git/nouveau/linux-2.6
~$ git remote add nouveau 
git://anongit.freedesktop.org/git/nouveau/linux-2.6
fatal: Not a git repository (or any parent up to mount parent )
Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).

What am I doing wrong?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #7 from Jonathan Nieder <jrnieder at gmail.com> 2012-06-13
15:49:56 PDT ---
> ~$ git remote add nouveau
git://anongit.freedesktop.org/git/nouveau/linux-2.6
> fatal: Not a git repository (or any parent up to mount parent )
> Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).
>
> What am I doing wrong?
My bad.  The step I left out is "cd linux".

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #8 from Gedalya <gedalya at gedalya.net> 2012-06-13 17:08:19
PDT ---
Created attachment 62998
  --> https://bugs.freedesktop.org/attachment.cgi?id=62998
Linux 3.5 - works

OK sorry that was really obvious.. :D

gdm3 starts fine with linux 3.5 and I can log in and use the desktop, so far I
see no obvious issue.

Guess I should try the 3.4.1 currently in experimental..

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #10 from Jonathan Nieder <jrnieder at gmail.com> 2012-06-13
18:43:24 PDT ---
Gedalya <gedalya at gedalya.net> wrote:
> OK, anything else I can do to help?
If you can find the fix by bisecting, that would make me very happy.

Note that this will be a little confusing.  When you say "git bisect
bad", that means the kernel did _not_ exhibit the problem.  When you
say "git bisect good", that means the kernel _did_ exhibit the
problem.  The "first bad commit" is the patch that fixed it. :)

It works like this:

    cd linux
    git bisect start v3.5-rc1 v3.3 -- drivers/gpu/drm/nouveau
    make deb-pkg; # maybe with -j4

That builds a version half-way between to test.  So:

    dpkg -i ../<name of package>; # as root
    reboot

    cd linux
    git bisect good; # if it crashed
    git bisect bad; # if it worked fine
    git bisect skip; # if some other bug makes it hard to test

Then another version to test will be automatically checked out
and the process continues.

Again, remember: good = crashy.

Finding the patch should take about 8 rounds.  If the gitk package is
installed, you can see the range with the fix narrowing by running
"git bisect visualize" at any step.  If you get bored before the
process finishes, "git bisect log" will list the results discovered so
far.  Even a few rounds can help a lot in narrowing down which patch
is responsible.

Then we should be better prepared to fix this in stable kernels.

Thanks,
Jonathan

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #11 from Gedalya <gedalya at gedalya.net> 2012-06-16 16:10:08
PDT ---
On 6/13/2012 9:43 PM, bugzilla-daemon at freedesktop.org
wrote:
> If you can find the fix by bisecting, that would make me very happy.
>
I hope I got it all right, I tried to be very careful...


$ git bisect bad
4cbb0f8d2b06c72aae3552ff1a0a57814c6ce7d2 is the first bad commit
commit 4cbb0f8d2b06c72aae3552ff1a0a57814c6ce7d2
Author: Ben Skeggs <bskeggs at redhat.com>
Date:   Mon Mar 12 15:23:44 2012 +1000

     drm/nvd0/disp: disconnect encoders before reprogramming them

     Signed-off-by: Ben Skeggs <bskeggs at redhat.com>

:040000 040000 4648bcf9d08c294fdb7ef76343e6f89e3cc2fe24 
2c31aa85ebf6c6f3b96fc6e91934a895bb42c0a3 M      drivers


$ git bisect log
# bad: [f8f5701bdaf9134b1f90e5044a82c66324d2073f] Linux 3.5-rc1
# good: [c16fa4f2ad19908a47c63d8fa436a1178438c7e7] Linux 3.3
git bisect start 'v3.5-rc1' 'v3.3' '--'
'drivers/gpu/drm/nouveau'
# bad: [4a206ffc0bfe8e8c3fc0468a052f5b0bb625a57b] drm/nouveau: oops, 
create m2mf for nvd9 too
git bisect bad 4a206ffc0bfe8e8c3fc0468a052f5b0bb625a57b
# good: [7d3a766b6aa4e293e72bfd6add477f05ac7fdf5a] drm/nouveau/pm: init 
only after display subsystem has been created
git bisect good 7d3a766b6aa4e293e72bfd6add477f05ac7fdf5a
# bad: [f1377998eede7a8caa124fcf6a589b02c9e2bac7] drm/nouveau: add 
userspace fallback hints.
git bisect bad f1377998eede7a8caa124fcf6a589b02c9e2bac7
# good: [c11dd0da5277596d0ccdccb745b273d69a94f2d7] drm/nouveau/pm: fix 
oops if chipset has no pm support at all
git bisect good c11dd0da5277596d0ccdccb745b273d69a94f2d7
# good: [6e83fda2c055f17780b2feef404f06803a49a261] drm/nvd0/disp: 
initial implementation of displayport
git bisect good 6e83fda2c055f17780b2feef404f06803a49a261
# good: [3488c57b983546e6bf4c9e0bfd0f7f2a1292267a] drm/nvd0/disp: move 
syncs/magic setup to or mode_set
git bisect good 3488c57b983546e6bf4c9e0bfd0f7f2a1292267a
# bad: [2f5394c3ed573de2ab18cdac503b8045cd16ac5e] drm/nouveau: map first 
page of mmio early and determine chipset earlier
git bisect bad 2f5394c3ed573de2ab18cdac503b8045cd16ac5e
# bad: [4cbb0f8d2b06c72aae3552ff1a0a57814c6ce7d2] drm/nvd0/disp: 
disconnect encoders before reprogramming them
git bisect bad 4cbb0f8d2b06c72aae3552ff1a0a57814c6ce7d2

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

Jonathan Nieder <jrnieder at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |skeggsb at gmail.com

--- Comment #12 from Jonathan Nieder <jrnieder at gmail.com> 2012-06-16
16:41:44 PDT ---
Adding Ben Skeggs to cc.  Ben, can you think of any reason not to include

 commit 4cbb0f8d2b06c72aae3552ff1a0a57814c6ce7d2
 Author: Ben Skeggs <bskeggs at redhat.com>
 Date:   Mon Mar 12 15:23:44 2012 +1000

      drm/nvd0/disp: disconnect encoders before reprogramming them

      Signed-off-by: Ben Skeggs <bskeggs at redhat.com>

in stable kernels?  Gedalya is finding it fixes a hang that can also
be experienced with 3.2.y and 3.3.y.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #13 from Jonathan Nieder <jrnieder at gmail.com> 2012-06-16
17:54:10 PDT ---
Created attachment 63122
  --> https://bugs.freedesktop.org/attachment.cgi?id=63122
drm/nvd0/disp: disconnect encoders before reprogramming them

Thanks again for your hard work tracking this down.

3.3.y (unlike 3.2.y) is not maintained any more, but the backport
to 3.3 is easier, so let's start there.

Please test the attached patch against the 3.3.y tree, like so:

  cd linux

  # fetch point releases
  git remote add stable \
    git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git
  git fetch stable

  # 3.3.y
  git checkout stable/linux-3.3.y
  make silentoldconfig
  make deb-pkg; # maybe with -j4
  dpkg -i ../<name of package>
  reboot

  # hopefully it reproduces the problem. So try the patch:
  cd linux
  git am -3sc /path/to/patch
  make deb-pkg; # maybe with -j4
  dpkg -i ../<name of package>
  reboot

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #14 from Gedalya <gedalya at gedalya.net> 2012-06-16 20:01:39
PDT ---
On 6/16/2012 8:54 PM, bugzilla-daemon at freedesktop.org
wrote:
> Thanks again for your hard work tracking this down.
Thank you!! You've been very helpful and this is a great learning 
experience!
> Please test the attached patch against the 3.3.y tree, like so:
>    # hopefully it reproduces the problem. So try the patch:
It did indeed.
>    cd linux
>    git am -3sc /path/to/patch
>    make deb-pkg; # maybe with -j4
>    dpkg -i ../<name of package>
>    reboot
>
So yea, it applied with no issues, compiled, and the computer seems to 
be quite usable, installed chromium etc. and so far so good.

This thing showed up, as before -
[   60.936501] colord[2769]: segfault at 8 ip 000000000040bc6d sp 
00007fff020a9110 error 4 in colord[400000+20000]
Not sure what it is, perhaps something else is wrong somewhere, perhaps 
unrelated.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #15 from Vova <vova7890 at mail.ru> 2012-08-18 17:42:16 UTC
---
What about crashes on linux 3.5.1 with applications uses 3d acceleration? I`m
booting with nouveau.noaccel=0. Gears sometime works, if enable kde effects -
driver is freeze.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=50571

Ilia Mirkin <imirkin at alum.mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #16 from Ilia Mirkin <imirkin at alum.mit.edu> ---
Sounds like the original issue is fixed. Vova, if you're still having
issues,
file a new bug describing your symptoms, take a look at
http://nouveau.freedesktop.org/wiki/Bugs/ for doing that effectively.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20130819/8155b2c1/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=50571

--- Comment #17 from Vova <vova7890 at mail.ru> ---
Yes, bug is fixed for me now.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20130822/6d55ebc7/attachment.html>