NFS client caching problems in 2.2.x (and proposal)

Frank van Maarseveen (F.vanMaarseveen@inter.NL.net)
Wed, 4 Aug 1999 22:50:05 +0200

I've been following some of the changes in NFS FS in 2.2.6-ac3,
2.2.10-ac?? and the 2.2.11-pre patches and it appears that a more
aggressive form of caching has been implemented. Unfortunately only
2.2.6-ac3 has an acceptable behavior (for my particular situation)
even though it has obvious bugs (reading the source and looking
at the OOPSes in /var/log/messages). The problem is that if one
mounts from the Internet one probably wants aggressive caching
but not in every situation: it must be possible to _disable_
these optimizations.

The heuristics implemented in the 2.2.10-ac11 and beyond are hard
coded and that is a _real_ problem:

fs/nfs/dir.c:nfs_neg_need_reval()
fs/nfs/dir.c:nfs_dentry_force_reval()

Propose:

/proc/fs/nfs/<mount-point>/acdirmin
/acdirmax
/acregmin
/acregmax
/nfs_neg_need_reval (?)
/nfs_dentry_force_reval (?)

A lot of NFS mount options can be made switchable this way without
having to remount the file system or to extend all the mount(2)
related stuff or figuring out the magic recipe for the heuristic
which satifies everyone. This recipe probably doesn't exist because
different situations demand different caching policies.

-- 
Frank


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/