XFS Scalability Improvement, Other File-System Enhancements Land In Linux 6.0

The XFS file-system may be getting up there in age but there are no signs of it slowing down but rather the opposite -- continued scalability work and performance optimizations -- as well as tacking on new features. With Linux 6.0, the mature XFS continues to age well. Separately, the EXT4 file-system updates have also landed in Linux 6.0.

First and foremost, XFS with Linux 6.0 brings a scalability improvement for large core count systems/servers. There is improved scalability of the XFS log by removing spinlocks and global synchronization points. The work by Darrick Wong was able to double transaction throughput without increasing CPU usage on large CPU count systems with these CIL optimizations.

XFS in Linux 6.0 also has various clean-ups, lockless lookups for the buffer cache to lessen the spinlock hammering, and various bug fixes from the past few weeks.

More details on the XFS changes for Linux 6.0 via this pull request with the material already having been merged. Looking forward, XFS maintainer Darrick Wong noted that he's going to be conducting a design review of XFS' online fsck handling with hopes of getting that merged soon.

Meanwhile Ted Ts'o has sent in the EXT4 updates for Linux 6.0. On the EXT4 side there are new ioctls to get/set the file-system UUID in the EXT4 superblock, improved performance of online resizing with BigAlloc enabled, and a number of bug fixes.