In addition to the Linux kernel's Zstd compression code being improved, the LZ4 compression code within the kernel tree is also seeing a separate set of enhancements.

Michal Wilczynski of Samsung sent out a set of patches on Friday to overhaul the in-kernel LZ4 code. Currently the LZ4 code in the kernel is effectively a fork of the upstream code that hasn't been updated since 2017~2018. Since then the LZ4 upstream code has seen bug fixes as well as performance optimizations.

The patch series by Wilczynski allows the LZ4 code to stop being forked for kernel use with instead being able to copy the upstream source code straight into the Linux kernel tree unmodified. The LZ4 code is properly vendored for the Linux kernel and allows moving forward to simply pull in the latest LZ4 compression/decompression code.

The LZ4 benefits of this clean-up and re-sync are quite impactful. The EROFS read-only file-system with LZ4 sees around 11% higher throughput with 4K clusters or up to 14% higher read throughput with 64K clusters. Other in-kernel LZ4 benchmarks show up to a 21% gain:

Those interested can see this patch series for the proposed big LZ4 update for in-kernel use.