FC6 x86_64 anaconda hangs on Dell XPS/Dimension 600 with nVidia RAID ("nodmraid" workaround)

[Stuart Levy]

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

[Jim Cornette]

>  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