Re: [PATCH] nfs: fix NULL pointer dereference in nfs4_fl_prepare_ds

[Jeff Layton <jlayton@redhat.com>]

On Wed, 25 Sep 2013 16:20:57 -0400
Jeff Layton <jlayton@redhat.com> wrote:

> In the event that there is an error connecting to a DS, we currently
> zero out the ds pointer and then try to dereference it to clear the
> NFS4DS_CONNECTING bit.
> 
> Fix this by clearing the bit unconditionally before zeroing out the
> pointer.
> 
> Cc: Andy Adamson <andros@netapp.com>
> Cc: <stable@vger.kernel.org> # v3.10+
> Signed-off-by: Jeff Layton <jlayton@redhat.com>
> ---
>  fs/nfs/nfs4filelayoutdev.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c
> index 95604f6..816cda8 100644
> --- a/fs/nfs/nfs4filelayoutdev.c
> +++ b/fs/nfs/nfs4filelayoutdev.c
> @@ -819,11 +819,11 @@ nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx)
>  		int err;
>  
>  		err = nfs4_ds_connect(s, ds);
> +		nfs4_clear_ds_conn_bit(ds);
>  		if (err) {
>  			nfs4_mark_deviceid_unavailable(devid);
>  			ds = NULL;
>  		}
> -		nfs4_clear_ds_conn_bit(ds);
>  	} else {
>  		/* Either ds is connected, or ds is NULL */
>  		nfs4_wait_ds_connect(ds);

Trond,

Please disregard this patch. There's another problem in here that we'll
need to fix too. Andy will be sending along a replacement patch later
that should fix this bug and the other problems too.

-- 
Jeff Layton <jlayton@redhat.com>