Greetings. There is an issue previously reported in at least the 2.4.18 kernel that affects people with Promise Ultra66 and Ultra100 IDE controllers; unsure about Ultra33 or other types. The issue is that when drives are attached to these controllers during a Linux install (Mandrake 8.2 in my case), the system will hang during a partition check on the attached drive. The system isn't completely dead; the CPU responds to things like numlock toggling. But it requires a reboot to get out of whatever loop the system appears to be stuck in.
The issue was discussed in this linux-kernel thread; follow the thread to see a reply by Alan Cox that at least indicates it was being looked into. I haven't found any additional info in the list archives. http://www.uwsg.iu.edu/hypermail/linux/kernel/0206.0/1264.html In addition, something appears to have changed in the version of the kernel in the cooker. Up until recently, I was able to use a 2.4.18 kernel on my file server when the raidtab was already present; somehow, this avoided the partition check on the drives in the raid set. But using this version of the kernel resulted in read data corruption, so I have been using the 2.4.8-34.1mdk (I think that's right) kernel. But the cooker version of the kernel that I downloaded over the weekend does NOT skip the partition check when the raidtab is present; it hung at the usual place at partition check. Note that I also had to install the cooker version of the modutils package, and I do not know if this had anything to do with the difference in operation of the previous 2.4.18 kernel I was using vs. the one in cooker. To me, this seems like a huge issue -- the data corruption issue was bad enough, but there is no possible way for me to use the version of the distro that's in cooker now, and perhaps in the 9.0 beta as well. I hope you can track down the bug(s) responsible for this behavior. I have full backups of my raid data set, so if there are changes implemented that should address these issues, I can test them out and report back the results. I'm not on the cooker mailing list, so feel free to email directly if re-testing is desired. Regards, - Leo Butler
