Domain 0 xen kernel has problems with scsi card on athlon system with stock FC4 xen kernel. The card is an older Adaptec 2940 which works flawlessly using the regular kernel. The system is booting from a software raid mirror on the local IDE disks. The only thing connected to the 2940 is an old Teac CDR drive. Everything works fine until the system tries to insmod aic7xxx. At that point it 'hangs', repeatidly resetting the scsi bus and sending abort commands. I've tried changing the IRQ settings for the card, disabeling xen (and the kernel) acpi support, booting into emergency mode and insmoding aic7xxx by hand.
from emergency mode, tried loading aic7xxx_old but it did the same thing. I get the first piece of messages where it "downloads sequencer code" etc. It's hanging just after that where it would normally start spitting out the individual device info. I moved the aic7xxx* modules in the xen0 module tree out of the way. Now I can get domain 0 to boot fine but have no CD Rom access. I'll try running an insmod of the module now that the machine is up. Then I can copy-paste the error here.
Created attachment 116239 [details] dmesg output After moving the modules out of the way and getting Domain-0 to boot I strace -ff -q -r -v -o /tmp/test insmod ./aic7xxx.ko. Attached is the dmesg output and the 'test' trace file though I don't know if it's very useful.
Created attachment 116240 [details] strace output did a kill -9 on the strace after letting it run for a while.
This is a really old kernel and I expect it is fixed with the new rawhide kernels --- certainly I've not seen any such SCSI problems on recent FC5test kernels. Please reopen the bug if you can reproduce with FC5test3 or rawhide.