FS-Cache Rewritten But Even Its Developers Are Hesitant About Landing It For Linux 5.9

FS-Cache provides the Linux kernel with a general purpose cache for network file-systems like NFS and AFS but also other special use-cases like ISO9660 file-systems. FS-Cache has been rewritten for better performance and reliability, among other benefits, and while it has been sent in as a pull request for Linux 5.9 even its own developers provide some caution over landing it this cycle.

FS-Cache has seen work to "massively overhaul" it with a variety of improvements. The new and improved FS-Cache will now use async direct I/O in place of snooping for updated pages that in turn means less virtual memory overhead. The new FS-Cache implementation has simpler object management, changes to object invalidation, and a variety of other work.

While all of the improvements sound good, currently this new FS-Cache disables the caching functionality for all file-systems besides AFS. Ceph, CIFS, NFS, and 9p are those affected with some of them having some patches available but not yet queued. So on these grounds alone it could get delayed from Linux 5.9 if it means multiple file-systems losing caching functionality.

Also unresolved are clearing file-system regression tests and some other low-level code matters.

There is the pull request to land the rewritten FS-Cache code so far for Linux 5.9, but as of writing Linus Torvalds hasn't yet provided his opinion on whether he will land it this cycle or not.