Red Hat Bugzilla – Bug 492640
Installation doesn't work onpart sda5 and sda6
Last modified: 2009-03-30 14:28:45 EDT
I've got a Dell Studio Hybrid for which I've pre-created the following partition scheme using the installation DVD, resizing an NTFS partition and creating an extended partition with two new partitions inside it (for swap and /) :
Disk /dev/sda: 320.0 GB, 320072933376 bytes
255 heads, 63 sectors/track, 38913 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0x80000000
Device Boot Start End Blocks Id System
/dev/sda1 1 8 64228+ de Dell Utility
/dev/sda2 9 1967 15728640 7 HPFS/NTFS
/dev/sda3 * 1967 16327 115353475+ 7 HPFS/NTFS
/dev/sda4 16328 38913 181422045 5 Extended
/dev/sda5 16328 16850 4200966 82 Linux swap / Solaris
/dev/sda6 16851 38913 177221016 83 Linux
My problem here is that I'm PXE booting to install Fedora 10 x86_64 on the system with the following relevant lines in the ks.cfg file :
# We have all partitions prepared, DO NOT TOUCH THEM!
part swap --onpart=sda5
part / --fstype=ext3 --onpart=sda6
The thing is that the installation fails everytime with errors about sda5 then sda6. When I check what's happening using the tty2 shell, I see that neither sda5 nor sda6 exist inside /dev at that point, only sda and sda1 to sda4. I tried hacking around the problem calling mknod from %pre, but at that point the mknod calls return that the files already exist. So I'm guessing that something is purging the sda* entries after %pre not re-creating them properly.
I've searched bugzilla for similar problems, but I only found ones that involved LVM, encrypted partitions etc. Nothing fancy here, I'm just asking anaconda to install on these 2 existing extended partitions and that's not working.
I've tested running anaconda from the Fedora 10 x86_64 DVD, and there when I choose to use a custom partition layout, I can select to use sda5 for swap and sda6 for / as they both appear fine in the graphical installer. At that point when I go to the console, I can see that all correct entries are in /dev.
Can you please give F11 Beta a try when it is released? We have made extensive changes to the storage code since F10 was released and it is hard to say right now whether your bug will still be present or not. Thanks.
Works fine with the current devel tree.
Nice improvements in devel btw, the NetworkManager part in anaconda no longer times out once before succeeding, so Fedora 11 will definitely be a huge improvement on this hardware! ;-)