Red Hat Bugzilla – Bug 154585
Fedora fails to write grub.conf properly,Windows will not boot after installation
Last modified: 2007-11-30 17:11:03 EST
Description of problem:
I've 2 hdd:
1st: is a SATA with one ntfs windows partition and some linux partitions (where
2nd: is a PATA with a fat32 partition (only storage data here, no OS at all)
I've done a clean FC4-T2 x86_64 install and when boot manager conf arrives, I've
choose to install GRUB to /dev/sda MBR (yes the 1st hdd, the SATA one).
The only one thing that I have changed is the driveorder FROM hda,sda TO sda,hda
I haven't touched the automatical recognized Windows partition.
When I rebooted after the installation Linux boot ok.
When I tried to boot Windows no OS is loaded, so I've entered Linux again,
edited my /etc/grub.conf and seen:
hd1 is mapped to PATA, so I changed to (hd0,0) and all works fine.
Version-Release number of selected component (if applicable):
Fedora Core 4 test 2
should have 2 hdd (in my case):
1st with a ntfs partition and windows installed and another partition with Linux
2nd with a fat32 partition and NO OS installed
Install Fedora Core 4 test 2 from DVD
When grub conf prompts change the order of the hard disk to put SATA (/dev/sda)
the first one and choose to install grub to the /dev/sda MBR
At first reboot try to boot Windows and will not boot until you change manually
Windows boot normally without manually change grub.conf
Created attachment 113053 [details]
Created attachment 113054 [details]
Created attachment 113055 [details]
buggy grub.conf created after installation
I installed FC4test2 on a 3 partition drive with FC3-release,swap and the
partition for FC4test2. I didn't touch the defaults for grub-config part of
anaconda and I got a grub which only boots FC4test2. I thought all partitions
with and OS would be included automatically. A bug or policy?
Bug still present in Fedora Core 4 test3.
Another small bug is that I have changed the name of the linux entry during
installation, and at the end of the installation I have discovered that Anaconda
changed to Linux (version of the kernel) that I don't want.
But of course the main bug is the most important.
I've replicated your scenario, and I don't think it is actually a bug. Changing
the "drive order" merely changes which disk anaconda thinks the BIOS will boot
off of, and thus to which disk we'll install a boot loader.
There's no way to know which of your other partitions have an OS installed --
that's why there's a screen to edit our guess!
Sorry I haven't understood why it isn't a bug.
I change the drive order ok, but I put sda as the hard disk that will have the
grub installed into its MBR, why anaconda thinks that the other hard disk will
contain another OS?
This bug is for me more serious when I have a correct grub.conf and anaconda
change it in a buggy one!
I've added this bug more than one month ago and none take care about it, just
one day after absolute devel freeze closed as "NOT A BUG" this is very strange
and seems to be a "closing all blocking bugs" also without fix them and without
a clear explanation...
As I said this IS a bug and fedora just don't want to fix.
I tried Kubuntu 5.10: same hardware, same hard disks, same partitions, same GRUB
v0.95 and Kubuntu was able to detect windows xp sp2 and write a fully functional