Livna crew, please read the conclusions. On Thu, Jan 04, 2007 at 12:29:41PM -0700, Charles Curley wrote: > On Thu, Jan 04, 2007 at 10:57:14AM -0800, oldman wrote: > > Charles Curley wrote: > > > if you get i586 as a response, you will have to download the i686 kernel > > and manually overwrite the existing kernel using: > > rpm -ivh --replacefiles --replacepkgs kernel > > I'll try this this evening. Yum is weird. I rebooted to a back kernel, erased the latest, and went to install it: [root@charlesc ~]# yum install $(locate kernel-2.6.18-1.2869.fc6.i6) Setting up Install Process Setting up repositories ... Reading repository metadata in from local files Parsing package install arguments Examining /var/ftp/pub/fc6.updates/i386/kernel-2.6.18-1.2869.fc6.i686.rpm: kernel - 2.6.18-1.2869.fc6.i686 /var/ftp/pub/fc6.updates/i386/kernel-2.6.18-1.2869.fc6.i686.rpm: does not update installed package. /var/ftp/pub/fc6.updates/i386/kernel-2.6.18-1.2869.fc6.i686.rpm: does not update installed package. Nothing to do I didn't ask it to update an installed package, I asked it to install a new one. Running "rpm -ivh" worked just fine. Running "yum install xorg-x11-drv-nvidia" (without rebooting to the new kernel) ran fine, except for the mysterious "FATAL: Module nvidia not found." message. I conjecture that comes from the RPM post-installation script trying to fire up the nvidia driver. Running it by hand produces: [root@charlesc ~]# service nvidia-glx start Checking for nvidia kernel module [FAILED] nvidia.ko module for 2.6.18-1.2868.fc6 kernel not found Disabling nvidia-glx driver [ OK ] Which is reasonable as I am on a kernel with no such module. On reboot, the system ran all the way to GDM. It painted about a quarter of the background screen, and then halted, leaving the rest of the screen white and locking up the rodent and the rest of X. There is a nice trace in /var/log/messages similar to the last one. I won't bore the list with it. I rebooted, and ran system-config-display. It went into a tizzy, making mouse movement very jerky. It then crashed, with a similar dump. I then compared xorg.conf with the backup made at installation time. They were identical. I rebooted to the old kernel, removed the nvidia packages, rebooted to the new kernel, and re-installed the nvidia modules. This time, no mystery message. I got the message about the nvidia module tainting the kernel. None the less, when I fired up X, I got the same symptoms: A partial loading of X and Gnome, the system going unresponsive, and then X crashing. Conclusions: * The kernel module is inserted when the rpm package is installed. This only works if the system is running the kernel for which the module is compiled. * It is not obvious that you have to be running the kernel for which a module is compiled while installing it. Installation may consist of dropping a file or two in the right place(s) and editing a few configuration files, neither of which requires a specific kernel. * The current error message is cryptic and user hostile. Recommendations: * If you must have a specific kernel, the time to test for it is before you start installing, not when you are almost done. This includes testing for architecture. * If the running kernel is not the one for which the module is compiled, refuse to install at all, and give the user a clear message saying why, and recommend the correct procedure. -- Charles Curley /"\ ASCII Ribbon Campaign Looking for fine software \ / Respect for open standards and/or writing? X No HTML/RTF in email http://www.charlescurley.com / \ No M$ Word docs in email Key fingerprint = CE5C 6645 A45A 64E4 94C0 809C FFF6 4C48 4ECD DFDB
Attachment:
pgppBAHvku325.pgp
Description: PGP signature
