Fast commits for ext4

Posted Jan 16, 2021 15:54 UTC (Sat) by mss (subscriber, #138799)
Parent article: Fast commits for ext4

> However, this change broke some existing tests that depend on fsync() working as a global barrier, so he backed it out.

It would be interesting to know how fsync() is implemented on other filesystems than ext*, whether it guarantees committing other files, too.

Looking at btrfs and xfs code it seems that they only sync the current inode when handling f(data)sync() calls so applications depending on fsync() being a FS-wide barrier are probably broken on these filesystems, too.