This is by design, these are called "stable pages". In some circumstances, the kernel has to ensure that pages aren't modified between initiating and completin

This is by design, these are called "stable pages". In some circumstances, the kernel has to ensure that pages aren't modified between initiating and completing the writeback for that specific page. Btrfs is a copy-on-write filesystem, so it ends up needing to use this guarantee more often than the others. This is something the btrfs developers are actively working on improving. Here's a good article describing this in a bit more detail: https://lwn.net/Articles/442355/ https://lwn.net/Articles/442355/ EDIT: Fix a typo