I've been trying to upgrade to FC6 x86_64 on a Dell XPS/Dimension 600. Both in install mode ("linux text") and in rescue mode ("linux rescue"), the symptom was that it'd boot the kernel normally, successfully probe for modules (usb_storage, ohci1394, sata_nv), successfully detect the graphics card, and then hang, with no error message.
Though the main screen was unresponsive, the Alt-F2 shell remained functional; it showed that anaconda was using a full CPU. Running strace (from the copy of FC4 already installed on the system) showed an endless loop of SIGSEGV / rt_sigaction / SIGSEGV etc.
After some digging through the anaconda code and running it in command-line debug mode (anaconda -d -C ...) from the Alt-F2 shell, it appears that the problem happens while trying to scan the system for RAID arrays, under
/usr/lib/booty/bootloaderInfo.py(444)createDriveList() dmsets = block.getRaidSets(drives)
I can step in pdb from within getRaidSets() to as far as:
/usr/lib64/python2.4/site-packages/block/__init__.py(138)getRaidSets() for rs in apply(c.get_raidsets, disks, {})
and then it gets into the SEGV loop, I suppose in the get_raidsets() method.
Workaround: add the "nodmraid" option to the boot command line, e.g. "linux rescue nodmraid" or "linux text nodmraid"
The system in question does have a hardware RAID known to the BIOS on the nVidia DXG051 motherboard, but it doesn't have a software RAID config -- /proc/mdstat doesn't show anything. The existing OS is just installed on ordinary DOS-labelled partitions, not using md/lvm/anything special.
There are three physical disks attached, and each is separately visible to the Linux kernel (sda, sdb, sdc), according to /proc/partitions.
dmraid -s -g *does* report two RAIDs -- I think one for each of sda and sdb. (How does it know?) I believe I configured the BIOS such that each of those two drives is a single-drive trivial RAID.
Hope this is useful. If I should provide more information, please let me know.
Stuart Levy
Hope this is useful. If I should provide more information, please let me know.
Stuart Levy
The information to me reads like a great diagnostic of problem and temporary solution. This bug sounds like it was mentioned on the test list and a bug was filed. I did not follow this issue but I believe I read messages that it was resolved for FC7.
You might want to check bugzilla for similar issues and add the information to the report.
Jim