Subject: Re: no FSCK required configuration.
To: Alex Nemirovsky <alexn@windriver.com>
From: der Mouse <mouse@Rodents.Montreal.QC.CA>
List: port-i386
Date: 02/16/2000 02:22:32
> Anyone know what I must to do to build a kernel which will not
> attempt to buffer the filesystem (ie. not do lazy writes).

Any kernel will do; just mount the filesystem(s) synchronous (-o sync,
or add "sync" to fstab).

> Filesystem performance is not a concern.

Good. :-)

> I want to allow power to be lost at any time, and not have to worry
> about filesystem inconsistences and as a result not have to do a fsck
> upon boot up.

This isn't possible with ffs.  There will still be windows between,
say, marking data blocks not-free in the bitmap and writing their
numbers into the inode (or indirect block(s)).  These windows will be
very small, and the userland process doing the I/O will be blocked
during them, but they're still going to be there, one way or another.
(This is not the only one; there are several others.  Writing directory
entries vs allocating inodes is another example, as is writing to the
three directories involved when doing a directory rename.)

I think LFS is supposed to be more crash-resilient; it may be up to the
task, though I don't feel competent to say for sure.  It's the only
NetBSD-supported filesystem that I think has any hope, at any rate,
though of course I could be wrong about that. :)

With the possible exception of NFS, which strikes me as cheating. :)

					der Mouse

			       mouse@rodents.montreal.qc.ca
		     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B