I sent a mail that make /etc/fstab clusterwide. Here is a brief
explanation of what i did.
(1) Introduce new file /etc/fstab.ssi.
(2) Made all mount point entires in /etc/fstab.ssi to use /devfs.
(3) mount uses my_getmntent to read the fstab entries. Reply only the
entries pertaining to that node. A simple 'strncmp()'
(4) swapon uses getmntent to read fstab entries and do a swap() on that.
Do a swap() only if the entry pertaining to this node. A simple
(5) Before mounting check for file system type. If supported by CFS, OR
the mount flags with MS_CFS.
I guess if the approach is fine we can replace cfs_mount and add
cfs_swapon command to the cluster tools with the above source. That
makes mounting easier and makes the startup script simpler in places of
mounting and adding swap
We can also remove the check for file system type inside the kernel for
setting the MS_CFS flags.