Hi Andy, thank you for the hint on the known bug, however currently I already run without any sata modules in initrd and use the /dev/hd* interface - and still get kernel Oopses (but don't have the complete system freezing anymore). When booting from DVD in recue mode, I have also been able to access the VIA SATA interface through /dev/hda?, does that mean I must not load the sata_via module in any way? So, I understand this applies to all sata_* drivers!? I have heard now, that this bug is fixed in the experimental (kraxel?) kernel 2.6.5 - at least in a way that the system run stable when using the old /dev/hd? interface. During the installation (incl. successful online update), my yast2 set-up got corrupted (rpms verify OK though). Now I can not start any yast2 modules anymore. Joel Wiramu Pauling describes a similar yet more problem in an earlier email to suse-amd64. Any explanation how this could have happened (and how it could be fixed)? What would be your recommendation? Format-n-reinstall maybe with different kernel-parameters? Or install experimental kernel. Which on-board SATA controller should I use? I heard that Promise only provides closed source modules? Gruss, Jan Andi Kleen wrote:
On Mon, May 17, 2004 at 12:09:25PM +0200, Jan Meyer wrote:
recently I tried to install Suse v9.1-AMD64 on my Athlon64 machine with an MSI K8T Neo-FIS2R (MS-6702) board. [Kernel was updated to 2.6.4-54.5]
There is a known bug in the 9.1 kernel that can cause problems when you have the sata_* driver loaded, but use the old style builtin IDE driver (which uses /dev/hd*). The sata driver can subtly corrupt the state of the IDE driver in some cases when it is loaded, which causes all kinds of strange problems later.
Workaround is to remove the sata driver from /etc/sysconfig/kernel:INITRD_MODULES and rebuild the initrd with /sbin/mkinitrd and reboot.
-Andi