You should look at how FSAL_GPFS does NFSv4 ACLs. NTFS ACLs will be pretty similar. Note that in the case of any operation that changes a directory (file create, rename, unlink), particularly rename, has unsolvable race conditions if Ganesha is relied on to do all the permission checking. Normally we defer permission checking of these operations to the underlying filesystem that can lock the parent directory and the directory entries being checked. This can be particularly problematical with the introduction of ACLs where a user might be allowed to delete /export/somedir/somefile so we pass the permission check but by the time we call the filesystem to do a rename that will delete somefile, the admin has deleted somefile and replaced it with another file. If that all transpires within the filesystem, the directory would be locked for the whole of the rename permission checks and actually performing all the steps for the rename, making the whole thing atomic. In that case, either the admin or the user go first and complete their operation. Frank