@GrahamcOfBorg test bcachefs

How is encrypted format broken for you? I just formatted with encryption, unlocked and mounted so I'm not sure what the problem is. Are you referring to #32279?

Yep, #32279 is pretty much what happened.

How well tested is this bcachefs version?
It gets stuck on fdatasync for me when trying to install NixOS on it.

I'm running it fine right now, albeit with some changes currently, manually made a patch for the latest kernel stable version (4.20.7) since the bcachefs repository is somewhat behind on the latest stable. Personally, I tested linux_testing_bcachefs before I made the PR, since bcachefs-tools and bcachefs were out of sync beforehand. Though unecrypted from bcachefs format, as stated above.

Is there a chance you can try to replicate koverstreet/bcachefs#16? If it doesn't occur without encryption you could cherry-pick #30333 (which isn't needed in initramfs and thus only necessary on the ISO used to install).

Note that the commit the kernel is picked from has since been amended as such

diff --git a/fs/bcachefs/btree_gc.h b/fs/bcachefs/btree_gc.h
index 1905acfa028a..8af5f841a537 100644
--- a/fs/bcachefs/btree_gc.h
+++ b/fs/bcachefs/btree_gc.h
@@ -109,7 +109,7 @@ static inline bool gc_visited(struct bch_fs *c, struct gc_pos pos)
 
        do {
                seq = read_seqcount_begin(&c->gc_pos_lock);
-               ret = gc_pos_cmp(pos, c->gc_pos) < 0;
+               ret = gc_pos_cmp(pos, c->gc_pos) <= 0;
        } while (read_seqcount_retry(&c->gc_pos_lock, seq));
 
        return ret;

I'll try tonight when I have more spare time, though the kernel issue doesn't really affect me right now since I have a patch containing all the commit diffs up to the previous revision (before the amend). As for bumping the revision on nixpkgs, I think it should be safe to assume that a new target for bcachefs{-tools} should come out relatively soon.

Looks like people should refrain from using zstd for now, see the linked issue. I thought it was just something weird about my system but looks like that is at fault - I don't think you have to reproduce now that we've found the factor.