Closed jcnelson closed 9 years ago
Actually, vdevfs will want to write data to the underlying filesystem, so if it crashes, the device nodes and information will be preserved. We'll mount a tmpfs on /dev/vdev instead, to store vdev's metadata.
On second thought, one workflow we'll want to support is: (1) start vdevd to populate /dev (2) mount /usr (3) mount vdevfs on /dev
This means vdevfs will need to plumb I/O through to the underlying filesystem.
Implemented as of 5fb243fd845da81099fdf28f572b58fafceb721f
Since vdevd now writes metadata and such to /dev/API/..., vdevfs needs to catch these writes and (preferably) store their contents in RAM.