F2FS & Btrfs Enjoy Some Nice Improvements With Linux 6.4
The Btrfs updates for the Linux 6.4 merge window were sent out today and include some performance improvements around directory logging. This change to avoid iterating over items and reducing locking contention during directory logging can yield a 4x lower fsync time.
When logging directory entries during one transaction, there is also now reduced locking of sub-volume trees that can yield a throughput improvement and lower latency during concurrent access to a Btrfs sub-volume.
Btrfs with Linux 6.4 is also seeing a clean-up to the I/O path, the scrub main loop has been completely rewritten, error handling improvements, and other changes. The improved Btrfs scrub code should be around 10% faster.
The full list of Btrfs feature patches for the Linux 6.4 merge window can be found via this pull request.
Also sent to Linus Torvalds today was the F2FS file-system driver updates. This updated F2FS code now allows non-power-of-two sized zones for zone devices, garbage collector code refactoring. support for I/O poll, and a variety of different bug fixes. Among the fixes are a possible system crash when lacking free space in certain conditions. See the full set of F2FS patches for Linux 6.4 via this pull request.
In related news on the file-system front, the FSCRYPT file-system encryption framework has landed various clean-ups and fixes, including preparations for FSCRYPT encryption with the CephFS file-system in a future kernel release.
13 Comments
