F2FS File-System Adding Device Aliasing Feature For Nifty Uses

[phoronix]

Phoronix: F2FS File-System Adding Device Aliasing Feature For Nifty Uses

The Flash-Friendly File-System (F2FS) is adding a device aliasing file feature that could appear in the upcoming Linux v6.13 merge window...

https://www.phoronix.com/news/F2FS-Device-Aliasing

[johanb]

So is this kind of a read/write version of squashfs, or am I missing something?

[bob l'eponge]

I think it's more like an optimized loop device. In a usual loop device, you are using a file for storage of a fake character device's content. So it means that each access to the mounted loop device goes to the VFS for the loop's filesystem, and then are mirrored via the loop driver to the VFS of the underlying filesystem with all the overhead (like block/page erasing, log, and so on). In short, it cost more than twice the device operations (since you need to account for the undelying VFS's overhead, like journaling/wear leveling and so on). If the FS on the loop system is journalized and specialized for flash system, then you'll pay the overhead twice.

In this case, the feature simply tell the underlying VFS to pass the operations directly to the actual hardware device. So you don't pay the overhead twice since management of the actual storage is left to the loop's device FS. On the down side, the area must be contiguous (so it can only happen at mount time, unlike a usual loop file that can be spread over the whole storage), it must be handled specifically (with additional boot scripts or configuration file) and depending on the actual software using the storage, can lead to increased wear on that specific area of storage (since the wear leveling of F2FS is disabled here).