Several people have confirmed a bug discovered in the Red Hat Linux 6.1 installation program (both GUI and text mode). When choosing the "Upgrade" option in installation, it gives an error message: "Error mounting ext2 filesystem on hdXX: Invalid argument. /dev/hdXX is not a ext2 partition". On the partitions it gives, it is indeed not a ext2 partition -- so the problem appears to be that it doesn't find the VALID ext2 partition on the drives. After the error message anaconda just sits there indefinitely with a "Searching for Red Hat Linux installations" dialogue. Upgrading seems almost impossible. All people who reproduced it were upgrading to 6.1 from 6.0 (not that it should matter). On my system I have 2 hard drives, with Linux on /dev/hda3 (hda1=windows). hdb is just an OS/2 (HPFS) partition. Note that Red Hat tried to treat a /dev/hdb partition as an ext2 partition. Suggestions: Try different partition arrangements with other filesystems to reproduce the bug. Add an installation command line option to SET where your ext2 partition is (in case Anaconda is unable to find it for some reason). At least then people like me (and the others) could get around this. ------- Additional Comments From 10/05/99 05:01 ------- I've had the same problem upgrading from a Redhat5.2/WinNT dual-boot system with the following partition layout: hda1 id 83 (ext2) Linux native (root partition) hda2 id 7 (OS/2 HPFS) WinNT NTFS hda3 id 5 (Extended) (contains other Linux paritions) The graphical upgrader fails when "Searching for Red Hat installations..." with the message "Error mounting ext2 filesystem on hda2: Invalid argument" The text-based upgrader fails with the same mesage as the graphical system: "Error mounting ext2 filesystem on hda2: Invalid argument". But using the text installer provides some Python stack information (I've had to retype this, so there may be a few minor typos in captialization, punctuation, and white space): Traceback (innermost last): File "/usr/bin/anaconda", line 225 in ? intf.run(todo, test=test) File "../../../RedHat/instimage/usr/lib/site-packages/text.py", line 1000 in run rc = apply(step[1](), step[2]) File "../../../RedHat/instimage/usr/lib/site-packages/text.py", line 222 in __call__ parts = todo.upgradeFindRoot() File "../../../RedHat/instimage/usr/lib/site-packages/todo.py", line 1130 in upgradeFindRoot isys.umount('/mnt/sysimage') File "/usr/lib/python1.5/site-packages/isys.py", line 8, in umount return _isys.umount(what) SystemError: (22, 'Invalid argument') ------- Additional Comments From 10/05/99 12:19 ------- It seems to be a bug in one of the python packages (_balkanmodule.so?) which the installation script is using to read the partition table. The type returned from the package is 2 for both NTFS and ext2 partitions!. The installation script looks for type 2 since it tries to mount all ext2 partitions. Of course it fails for NTFS partitions. Note that HPFS partitions look the same as NTFS in partition table, so it may happen with HPFS as well. Bolek bolek ------- Additional Comments From 10/07/99 03:32 ------- I have the same error on this simple disk: /dev/hda3 as ext2 partition /dev/hda1 as fat32 partition (WIN98) Regards ------- Additional Comments From 10/10/99 13:40 ------- It seems like Redhat 6.2 doesn't check if the Linux partition is on any partition with a number higher than 4. My Redhat 6.0 is installed on /dev/hda6 and 6.1 upgrade doesn't find it. After installing a test Redhat 6.0 to /dev/hda3 the installer upgrades this one. What do people do who have two Redhat versions on their system? ------- Additional Comments From 10/10/99 14:06 ------- I've also been unable to perform an upgrade from 6.0 to 6.1 on ANY (2) of the systems I've tried. Both systems had both IDE and SCSI disks with the Linux installation on the SCSI disks. Python traceback and all other symtoms are identical to those already discussed.
Same problem. I have NTFS on /dev/sda and /dev/sdb1. My ext2 is on /dev/sdb3. Cannot upgrade. Would be nice if I could indicate to the install program that my ext2 is on /dev/sdb3. ------- Additional Comments From 10/14/99 03:01 ------- Not Only the missing break for NTFS/HPFS in anaconda/balkan/dos.c causes the upgrade to abort. This source file does not know that Microsoft invented a new type for extended partitions. These days we not only have 0x5 but also 0xf (which Linux fdisk calls Win95 Ext'd (LBA). The RedHat installation mistakenly thinks this is a DOS partition. Instead all checks made with DOSP_TYPE_EXTENDED/5 must be extended to also include 0xf as a type for exended partitions. To my knowledge this new partition type is used for harddisk greater than 8gb and not only Windows uses it but also Partition Magic from Powerquest. So as a workaround one not only has to change the id of NTFS/HPFS partitions (of course only temporarily to for example 0xa7) but also must set extended partitions with type 0xf to have type 0x5.
Although this may be overkill, I will add my two cents. I two have found this bug upgrading 6.0 to 6.1. I recieve the same error message(s) in both text and graphic mode. /dev/sda1 = WinNT /dev/sdb1 = Linux (REDHAT 6.1) Just wanted to make sure it was known that SCSI disks are effected as well (although it might have been obvious, just making sure).
Just to confirm that a workaround reported 10/14/99 03:01 works. I booted from my existing install using linux single, ran fdisk and changed my NTFS partiitions to type a7. Rebooted to the 6.1 CD and ran the upgrade without a problem. Rebooted to 6.1 linux single and used fdisk to change the partition types for NTFS back to normal. Edited my lilo.conf (as it lost the dual boot config) and re-ran /sbin/lilo. Rebooted and all works fine now. Thanks for the workaround!
We will be releasing an errata update image which solves this problem the week of October 18.
I have post another bug report (with the new installation images #6328). I have two systems, OS/2 (Warp 4) and Linux (6.0): sda5, extended partition OS/2 sda6, extended partition OS/2 sda7, extended, ext2 sda8, extended, HPFS sda9, extended, ext2 ... The first partition that the installer tries mount is a HPFS partition and the installer crash. The errata uptade image does not fix this bug. I read that it be possible to set the ext2 partition in a command line option. Can you explain, please? Best regards, JKB