Re: [PATCH 0/3] Data in direntry (dirdata) feature

["Theodore Tso" <tytso@mit.edu>]

On Sat, Apr 18, 2026 at 11:24:32PM +0100, Artem Blagodarenko wrote:
> There is now a way to set dirdata. Following the decisions from [1]
> the patches provided change how the encryption + casefold
> hash is stored. If the dirdata feature is enabled,
> the hash is stored as dirdata. Therefore, to test dirdata, it is
> sufficient to test the combination of encryption + casefold + dirdata.

I'm not seeing that in the patches that was sent out to the list last
week.  Where is that?

I traced all of the places where ext4_insert_dentry_data() and
ext4_dirdata_set() and I don't see *anything* where dirdata was
stored, including the fscrypt + casefold hash.   What am I missing?

It *really* would be a good idea if the e2fsprogs patches included a
way to list and set the dirdata using debugfs.  That way we could
easily verify that dirdata field was getting set when you expected it
to be.

> ext4/064: encryption + casefold (WITHOUT dirdata)
> 
> This test verifies that files created in directories with both
> encryption and case-insensitive (casefold) attributes work correctly.

Yes, but that doesn't actually verify that the dirdata field was set
when you expected it to be.  Just that it works correctly....

> LUFID is an optimization that allows Lustre to store file identifiers
> efficiently directly in directory entries, avoiding additional I/O
> operations to look them up separately.

Oh, I see.  So this is for readdir(), right?

> I am open to adding mechanisms to access LUFID outside the Lustre
> filesystem. However, since: 1) dirdata is tested using the encryption +
> casefold features, and LUFID is primarily useful within the Lustre filesystem
> and is already tested there, I would prefer to rely on testing LUFID
> within Lustre FS.

Well, the one advantage of having a way to set and get LUFID would be
if you ever wanted to ressurrect the userspace Lustre server[1].  :-)

[1]  https://wiki.lustre.org/images/5/56/LUG08-Lustre-uOSS.pdf

And I *do* think it would be useful to have a way to set and get the
LUFID using debugfs.

Cheers,

						- Ted