On Mon, 2020-11-16 at 22:18 -0500, J. Bruce Fields wrote:
> From: "J. Bruce Fields" <bfields@xxxxxxxxxx>
>
> Currently when knfsd re-exports an NFS filesystem, it uses the ctime as
> the change attribute. But obviously we have a real change
> attribute--the one that was returned from the original server. We
> should just use that.
>
> Signed-off-by: J. Bruce Fields <bfields@xxxxxxxxxx>
> ---
> fs/nfs/super.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/nfs/super.c b/fs/nfs/super.c
> index 4034102010f0..ca85f81d1b9e 100644
> --- a/fs/nfs/super.c
> +++ b/fs/nfs/super.c
> @@ -1045,6 +1045,7 @@ static void nfs_fill_super(struct super_block *sb, struct nfs_fs_context *ctx)
> } else {
> sb->s_time_min = S64_MIN;
> sb->s_time_max = S64_MAX;
> + sb->s_flags |= SB_I_VERSION;
> }
>
>
> sb->s_magic = NFS_SUPER_MAGIC;
I don't think we want this change. This will make file_update_time
attempt to bump the i_version field itself using the routines in
iversion.h. This will almost certainly do the wrong thing.
--
Jeff Layton <jlayton@xxxxxxxxxx>
