Re: [PATCH v3 08/16] xfs: Report case sensitivity in fileattr_get

["Darrick J. Wong" <djwong@kernel.org>]

On Mon, Jan 12, 2026 at 12:46:21PM -0500, Chuck Lever wrote:
> From: Chuck Lever <chuck.lever@oracle.com>
> 
> Upper layers such as NFSD need to query whether a filesystem is
> case-sensitive. Populate the case_insensitive and case_preserving
> fields in xfs_fileattr_get(). XFS always preserves case. XFS is
> case-sensitive by default, but supports ASCII case-insensitive
> lookups when formatted with the ASCIICI feature flag.
> 
> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Well as a pure binary statement of xfs' capabilities, this is correct so:
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

[add ngompa]

But the next obvious question I would have as a userspace programmer is
"case insensitive how, exactly?", which was the topic of the previous
revision.  Somewhere out there there's a program / emulation layer that
will want to know the exact transformation when doing a non-memcmp
lookup.  Probably Winderz casefolding has behaved differently every
release since the start of NTFS, etc.

I don't know how to solve that, other than the fs compiles its
case-flattening code into a bpf program and exports that where someone
can read() it and run/analyze/reverse engineer it.  But ugh, Linus is
right that this area is a mess. :/

--D

> ---
>  fs/xfs/xfs_ioctl.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> index 59eaad774371..97314fcb7732 100644
> --- a/fs/xfs/xfs_ioctl.c
> +++ b/fs/xfs/xfs_ioctl.c
> @@ -516,6 +516,13 @@ xfs_fileattr_get(
>  	xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
>  	xfs_iunlock(ip, XFS_ILOCK_SHARED);
>  
> +	/*
> +	 * XFS is case-sensitive by default, but can be formatted with
> +	 * ASCII case-insensitive mode enabled.
> +	 */
> +	fa->case_insensitive = xfs_has_asciici(ip->i_mount);
> +	fa->case_preserving = true;
> +
>  	return 0;
>  }
>  
> -- 
> 2.52.0
> 
>