I installed RH7.1/XFS on a system with a root aic7xxx scsi
drive and a DAC960. I did XFS for all partitions. The install
went smoothly. Everything worked great. Then, I rebooted
(no reason, a mistake actually...)
When it came back up, /etc/sysconfig/hwconf was replaced with
all zeros, the XFS log is corrupted, and it immediately gives
Oopses in nfsd. Not sure what happened, but these look like\
problems addressed earlier, and probably fixed in CVS.
A few questions:
Why do I get zero-filled files on a crash, but I never see
them on an SGI crash? Also, I never see it with ext2, which
probably leaves the previous contents for unsynced files
insteds of zeros. It would be nice if unsynced XFS would
leave the previous contents for unsynced files, and it would
also be good to store a list of suspect files on reboot.
It seems like almost half the questions stem from devfs
differences, which probably also is the culprit for install
problems with DAC960 root. Maybe devfs should be an optional
feature.
Thanks, Joe Krahn
|