Description of problem: Installing FC6 with XFS instead of ext3 yeilds filesystem issues after the install. Version-Release number of selected component (if applicable): FC6, base kernel & xfsprogs. How reproducible: Install FC6 system with defaults, _except_ substitute xfs for ext3 as the file system on all non-/boot partitions. Actual results: Post-mounting of filesystems, kernel messages spew across the screen: <1>I/O error in filesystem("dm-0") meta-data dev dm-0 block 0x2abb8000000 ("xfs_trans_read_buf") error 5 buf count 4096 <6>attempt to access beyond end of device <6>dm-0: rw=0, want=2936549670920, limit=12255323 Repeated multitudes of times. These messages were captured by rebooting installer in rescue mode. Note that the filesystems appear to mount cleanly, and minor operations (e.g. "cat etc/fstab") work. The end system is unusable :( Expected results: Spiffy new FC6 system with xfs goodness. Additional info: Note this does not appear to happen on FC5 systems with lvm/xfs upgraded to FC6.
On the fourth attempt, I rebooted into the installers rescue mode and xfs_check'ed the three xfs filesystems (/, /tmp. /home). The root fs was seriously corrupted; xfs_repair was unable to fix. Recreating these filesystems from within the rescue shell with mkfs.xfs (sticking with all defaults except labeling (-L) each fs) and reinstalling w/o reformatting yielded a working FC6 system. I don't know if this was a one-off (in which case I'll be a touch embarassed) or if this was indeed a bug.
problems w/ the mkfs command in anaconda for some very strange reason... duping to bug w/ more info. *** This bug has been marked as a duplicate of 212201 ***