Comment 10 for bug 235262

Revision history for this message
rannsaicher (rannsaicher) wrote :

I HAVE A FIX
Following the clue i found above - that vermagic is the same for all dkms installs, i decided to reinstall the ati_8.6 driver whilst running under the 2.6.24-19 kernel as it seemed maybe the ati installer generated module was carrying its vermagic and other symbols etc from the -16 kernel definitions used in the compile, onwards. Another clue was the Kernel preparation unneccesary message during dkms install.
HERES THE STEPS
Commands prefaced by $
I was stuck on 2.6.24-16 kernel so I booted into 2.6.24-16 and used dkms to remove the kernel module from the 2.6.24-19 kernel (this step probably not required, as later on the install seemed to clean fglrx out of my dkms tree)
in a terminal window.
$dkms status
to show what versions and kernels your modules are installed against.
$sudo dkms remove -m fglrx -v 8.501 -k 2.6.24-19
...removes the 8.501 version from the -19 kernel.

I booted into 2.6.24-19 recovery mode ( pick it from the grub menu), and now as root, cd to where the ati 8.6 installer is located and run the installer
$sudo sh ati-driver-installer-8-6-x86.x86_64.run --buildpkg Ubuntu/hardy
after the installer completes, install the generated packages
$sudo dpkg -i *.deb
during the install of packages dkms is invoked, and at this point it cleaned out all fglrx modules ( including the one on the -16 kernel that was my lifeline !)
check the kernel module is available
$dkms status
which gives me this output;
fglrx, 8.501, 2.6.24-19-386, i686: installed
confirming the kernel module fglrx is available for the -19 kernel
$reboot
and it came up on -19 kernel, allowing me to log in and voila- no white screen anymore.

Wonder if theres an issue with dkms?

Enjoy.