#613747 xserver-xorg-video-radeon: kernel oops observed on init start/shutdown with KMS enabled

Package:
xserver-xorg-video-radeon
Source:
xserver-xorg-video-ati
Description:
X.Org X server -- AMD/ATI Radeon display driver
Submitter:
Martin Strauss
Date:
2011-08-03 12:54:04 UTC
Severity:
important
#613747#5
Date:
2011-02-16 23:02:39 UTC
From:
To:
Hello,
I installed debian stable squeeze on my laptop lenovo T60 with a Radeon
Mobility X1400 graphic chip.
Unfortunally I observed various kernel oops at boot time (init process)
and sometimes also at shutdown. The kernel stacktrace seems to arbitrary
(ie ext4 code, etc).
In seldom cases the screen does switch from console to a black screen
with some lines on it and then freezes.
If X Server comes up with gdm it seems to run stable.

I installed the linux-firmware-nonfree package which did not help,
Fortunally booting with init=/bin/bash always worked so I was able to
recover.

After some internet research I disabled KMS in
/etc/modprobe.d/radeon-kms.conf
which made the system rock solid.

After reading the other bug reports I had a look whether linux-firmware-free
was installed and it was missing. So I had another try with KMS enabled
after installing linux-firmware-free and booted 5 times ... it worked
(Usually it crashed ever second time or so).
I removed linux-firmware-free and rebooted again (with kms) ... no
crashes ...

So either the system updates of the last to weeks made a difference or
i was just lucky today or having installed linux-firmware-free once is
just what it needs.

As the default installation produced an unusable installation which only
an advanced linux user would be able to repair I classified this report
as Important.

For my self it seems to work (or today is my lucky day).
As I don't trust my luck ...

Is it possible that the missing of linux-firmware-free produces kernel oops?

If interested and if someone can give some advice howto
I could spend the time to do some research ...
currently I have no idea how to reconstruct the effect.

sincerly
Martin Strauss

#613747#12
Date:
2011-03-27 15:51:21 UTC
From:
To:
Hello,
I had the hope that the reported unstablities would no longer occure.
Unfortunally that's not true, it just seems to occure less often.
So even the default amd64 installation with firmware-linux-free  and
firmware-linux-nonfree and kms enabled makes the system unstable.
As it seems the system hangs between the start of the init process
and the X login, usally the display is frozen with some console output
with some horizontal pixel distortions.
What would be the next steps that I could to help to make the kms
enabled mode stable ?
sincerly
Martin Strauß

#613747#17
Date:
2011-08-03 12:51:01 UTC
From:
To:
I've observed frequent but inconsistent kernel oops on boot and shutdown of Debian Squeeze on 5 of 7 tested HP Compaq dc7800 machines.

Running 'lspci | grep VGA' yields:

  01:00.0 VGA compatible controller: ATI Technologies Inc RV530 [Radeon X1600]

Running 'lsmod | grep radeon' yields:

  radeon                574844  2
  ttm                    40146  1 radeon
  drm_kms_helper         20369  1 radeon
  drm                   142327  4 radeon,ttm,drm_kms_helper
  i2c_algo_bit            4225  1 radeon
  i2c_core               15819  4 radeon,drm_kms_helper,drm,i2c_algo_bit

Adding 'blacklist radeon' to '/etc/modprobe.d/blacklist.conf' changes the output of lsmod upon reboot for radeon from:

    radeon                574844  2

to:

    radeon                574844  1

A custom minimal install of Debian without X yielded:

      radeon                574844  0

In the last two cases (system with 'blacklist radeon' added and system with minimal install without X), I have been unable to reproduce a kernel oops testing on multiple machines across many reboots.

Gnome still works with the addition of the blacklist. However, I have observed tty switching issues (i.e. if I switch to tty1 then back to tty7 then try to go back to another tty I can only get tty7 to come up).