I have a Gentoo Linux 1.2 system with xfs. When the power fails and the computer restarts, the update-modules script makes a corrupted /etc/modules.conf file. Why?, I suppose xfs is robust but, is possible that the update-modules script fails during the xfs partition checking (at the mount moment)?. I have decided to comment the lines of the update-modules at the boot rc script to fix this problem for now.
What do you mean by it corrupts it ?
It contains trash intead of a plain text file. This makes that the kernel modules don't load at the startup process.
Ok, try to look at the files in /etc/modules.d/, as these are used to generate /etc/modules.conf. I am assuming one of these are corrupted.
No, the files under /etc/modules.d/ are ok (I checked them for the first time). I think it is a problem with the update-modules running at boot time with a xfs checking.
What exactly do you mean by "boot time with a xfs checking" ?
XFS is pretty icky. Try another FS and see if you can duplicate it. I've never seen a decently stable XFS release or CVS. =(
Any new developments ?
I can't duplicate this, neither can anyone that uses XFS... if we can get it to duplicate elsewhere, lemme know. sounds like your FS got corrupted somewhere and you just need to dd the disk and start over. =( Sorry...
I started having the same problem w/ reiserfs on the 2.4.2x kernels (as late as of 2.4.23). I found that removing the Crypto API modules from being compiled in-kernel solved my stability and disk corruption problems. Details: currently kernel 2.4.23 on dual Xeon PII 400MHz w/ 640MB RAM; running ide on mainboard and PCI Promise U66 card; reiserfs on non-boot partitions; no sound; ATI Rage AGP video