OK, I'm *not* dual booting. Just FC2. Whatever they've done with the 2.6 geometry code detection/translation/hashings has FUBAR'd this puppy beyond use. Apprently this issue isn't being seen by users of the current 2.6 kernel. This looks to have been fixed months ago and didn't make it into FC2 - yet. See article at the end here for details. Or I'm just not seeing what I think I'm finding... (wouldn't be the first time)
P-III 500 - 1999 BIOS (flashed to a mid 2000 version later on) - 2 Hard drives, 10GB and 80GB.
/dev/hda is 10GB drive and is dedicated to /boot and swap - nothing else.
/dev/hdb is 80GB and has everything else on it.
Brand new FC2 install, grub is installed to MBR on hda. On boot all I get is a grub shell. Tossed a few geo commands to see what's going on:
grub>geometry (hd0)
drive 0x80: C/H/S = 512/32/63, The number of sectors = 1032192, LBA
Partition num : 0, Filesystem type is ext2fs, partition type 0x83
Partition num : 1,
Error 18: Selected cylinder exceeds maximum supported by BIOS
After a moment of WTF?? I check the BIOS settings. Sure enough - they're now set to: 512/32/63. Something on boot up is resetting the BIOS to values that don't make any sense. So, I re-detect the drive, save the correct geometry into BIOS and boot again. Same results.
::SIGH::
Try again at the grub shell to see what's up, but this time feeding grub the correct geometry.
grub> geometry (hd0) 1024 768 63
Same exact dang-blasted frustrating result as detailed above.
Just to be sure I wasen't be'n prank'd, pulled the GPG key from fedora.redhat.com - again, re-verified that the md5sums I have are legit. Re-verified the iso md5s - even did the mediacheck after burning - twice - all OK. It just flat out dosen't work with large drives on an older BIOS.
FWIW - this same hardware was working perfectly under FC1 - 2.4 kernel yum'd to latest.
I'm amazed this was released like this. Coupled with the dual boot "Destroy your W2K/XP install" feature - this is disasterous. I mean, I know we're the 'development' lab-rats for Redhat - but come on. Most labs at least feed and water their test subjects - maybe even a T-shirt, or some cedar shreddings...
Here's an article
Google Cache Article that describes what appears to be this same problem and the fix for it that was submitted to the kernel devel folks some time ago.
************************
And my fix?? Well - apparently not really a 'fix' and certainly not what was listed above as a 'fix', apparnlty just a work-around. After completing what appearsto be a normal and successful, clean install - re-boot the sysetm. Same problem rears it's head as expected. Now, go into BIOS setup and after re-auto-detecting the boot drive's geometry, I choose an option that did *NOT* use LBA. Yes - redetect and choose *AFTER* the succesful insatll w/ drives set to use LBA, but choose a non-LBA geometry.
FC2 came up all smiley and functional. I'm have'n a blast now!!
Just curious if this behavior really is related to the bug/fix mentioned in that google article or not.