promac at gmail
May 26, 2010, 5:29 PM
Post #3 of 3
On Wed, May 26, 2010 at 7:47 PM, Kirk Bocek <t004 [at] kbocek> wrote:
> John Pilkington wrote:
>> I've just had what seemed to be an nvidia problem in CentOS 5.5, but it
>> turned out that the default boot kernel hadn't been updated from 194 to
>> 194.3.1 Since I only had kmdls for the updated kernel installed everything
>> became rather difficult.
>> I remember thet long ago I used to disable a newly installed kernel until
>> I had a full set of kmdls. Now I usually delay installation until the kmdls
>> are available, and I had thought that the new kernel had then automatically
>> become the default. Apparently not.
>> If this isn't just my past coming back to haunt me this note may sound a
>> useful alarm.
> I've upgraded four hosts from 5.4 to 5.5. In every case the 194.3.1 kernel
> was installed and made the default kernel.
It depends on what you have in /etc/sysconfig/kernel:
[cascavel:~/SRPMS/atrpms] more /etc/sysconfig/kernel
# UPDATEDEFAULT specifies if new-kernel-pkg should make
# new kernels the default
Also note that if a new nvidia version is installed simultaneously with a
new kernel, the old kernel may not have a corresponding kmdl for the new
nvidia version and x will not start, unless the boot is through the new
Another point is that we are using a new symbolic link for the current
nvidia version, and the script nvidia.py should be used to fix previously
installed versions (Fedora and RHEL use the same scheme).
Paulo Roma Cavalcanti
LCG - UFRJ