ReiserFS Without Big Kernel Lock

Aug 03, 2009

Even though reiserfs belongs to the veterans of journaling filesystems and established ext3 and ext4 as standard Linux filesystems, a new patch serves to markedly improve it.

The Kill-bkl patch, which Frederic Weisbecker wrote and is now on the kernel mailing list for kernel integration, replaces the previous reiserfs locking scheme that was based on a big kernel lock (bkl). The new write locking system is more granular and works on superblocks, which essentially provides a separate lock for each reiserfs, thus providing a measurable performance advantage for large installations.

According to Weisbecker's initial measurements, single-thread and multi-thread processes profit from the more granular locking scheme. The patch to the current reiserfs is available as a diff file or in the corresponding git tree.

Related content

comments powered by Disqus