Before getting too deep into cachefs issues, bear in mind that almost _all_ the existing NFSv3 code is pretty rotten and breaks horribly under high load situations with a good chance of file corruption if multiple NFS clients attempt writes.
This gets even worse if the source filesystem is being shared via Samba (or any other protocol) and worse still if the source filesystem is a clustered one such as GFS.
It may well be considering building this around some alternative to NFS, unless the existing server/client code is rewritten from the ground up.
(In particular the linux NFS server needs to come out of kernel space and back into userspace in order to play nicely with other protocols (such as samba, which didn't exist when the server migrated into the kernel 17 years ago).
v4 attempts to address a lot of the issues but I find it's still not production ready - particularly in a high-availability environment.)
AB -- Linux-cachefs mailing list Linux-cachefs@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cachefs
