Add support for `bcachefs`

[Schweber]

The manual states that the -d option only works on btrfs and xfs.

It is my understanding (please correct me if i'm wrong) that all filesystems with reflink can be deduplicated. This would include bcachefs, zfs and f2fs.

If this is indeed the case, i kindly suggest to support them as well.

[ErrorNoInternet]

They need to support a specific ioctl, not just reflink. But if they support reflinks then it should be fairly straightforward for them to implement the ioctl.

ZFS for example: openzfs/zfs#15393

[Schweber]

I see. Thank you for the explaination and sorry for my ignorance.

[trougnouf]

It looks like bcachefs supports the necessary FIDEDUPERANGE ioctl (per koverstreet/bcachefs#622 ).

[ticpu]

Shouldn't this issue be renamed to Add support for bcachefs to keep it within a reasonable and achievable scope?

[Schweber]

You got it

[ticpu]

@Schweber did you try duperemove on bcachefs recently? I think it works much better since reflink support was improved in 6.13

I could do a complete run on my Unity folder and saved about 6GB for 39GB.

[Schweber]

No, i didn't try it as it's no officially supported.

[brvtalcake]

After looking to Bcachefs' kernel code, and thanks to Linux kernel documentation for VFS, I can tell (nearly) for sure that (at least) trunk Bcachefs implements what it needs to support FIDEDUPERANGE. That being said, the problem I personally encountered when trying duperemove on my 4 partitions Bcachefs (as my rootfs, yes) (I'm a fully consenting masochist) was when it tried to get the filesystem's UUID. Maybe Bcachefs needs the same kind of code that Btrfs has in btrfs-util.c for multi-devices filesystems ?

I didn't investigate any further (and I didn't really understand why it needed the UUID), but I could as soon as I have time, if you wish / if needed / if useful, for that matter

EDIT, just for reference and future wanderers : if I'm not mistaken, -dd causes duperemove to ignore the filesystem type check (or maybe this issue was created before -dd was a thing ?)

[ticpu]

@brvtalcake did you try master yet? With #364 I think the issue you mention is fixed.

[brvtalcake]

@brvtalcake did you try master yet? With #364 I think the issue you mention is fixed.

@ticpu thank you very much, I didn't see that ! And sorry for the very late reply, I will try the master branch this afternoon (until now I was using only the version provided by Gentoo)