Re: [PATCH] nfs: explicitly reject LOCK_MAND flock() requests

[Jeff Layton <jlayton@redhat.com>]

On Mon, 23 Jul 2012 15:46:23 -0400
Jeff Layton <jlayton@redhat.com> wrote:

> We have no mechanism to emulate LOCK_MAND locks on NFSv4, so explicitly
> return -EINVAL if someone requests it.
> 
> Signed-off-by: Jeff Layton <jlayton@redhat.com>
> ---
>  fs/nfs/file.c |    9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/fs/nfs/file.c b/fs/nfs/file.c
> index 61d3670..15f4bbb 100644
> --- a/fs/nfs/file.c
> +++ b/fs/nfs/file.c
> @@ -834,6 +834,15 @@ static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
>  	if (!(fl->fl_flags & FL_FLOCK))
>  		return -ENOLCK;
>  
> +	/*
> +	 * The NFSv4 protocol doesn't support LOCK_MAND, which is not part of
> +	 * any standard. In principle we might be able to support LOCK_MAND
> +	 * on NFSv2/3 since NLMv3/4 support DOS share modes, but for now the
> +	 * NFS code is not set up for it.
> +	 */
> +	if (fl->fl_type & LOCK_MAND)
> +		return -EINVAL;
> +
>  	if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK)
>  		is_local = 1;
>  

Hmm...it looks like GFS2 does a similar check and returns -EOPNOTSUPP.
Should we do the same here instead of -EINVAL?

-- 
Jeff Layton <jlayton@redhat.com>