OK – I’ll take a look.
Thanks
→Andy
On 3/15/17, 3:09 PM, "Trond Myklebust" <trondmy@xxxxxxxxxxxxxxx> wrote:
On Wed, 2017-03-15 at 11:18 -0400, andros@xxxxxxxxxx wrote:
> From: Andy Adamson <andros@xxxxxxxxxx>
>
> Otherwise no slots could be available for the potential
> pnfs_layout_process getdeviceinfo call and the client hangs on pNFS
> I/O.
>
> This occurs in testing against the pynfs pNFS server where the
> the on-wire reply highest_slotid and slot id are zero, and the
> target high slot id is 8 (negotiated in CREATE_SESSION).
>
> The internal fore channel slot table max_slotid, the maximum allowed
> table slotid value, has been reduced via nfs41_set_max_slotid_locked
> from 8 to 1. Thus there is one slot (slotid 0) available for use
> but
> it has not been freed by layoutget proir to the getdeviceinfo
> request.
>
> So the layout get call is waiting for the getdeviceinfo call to
> complete,
> and the getdeviceinfo call is waiting for layoutget to free the slot.
>
> Signed-off-by: Andy Adamson <andros@xxxxxxxxxx>
> ---
> fs/nfs/nfs4proc.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
> index 1b18368..8ca92d0 100644
> --- a/fs/nfs/nfs4proc.c
> +++ b/fs/nfs/nfs4proc.c
> @@ -8504,10 +8504,10 @@ nfs4_proc_layoutget(struct nfs4_layoutget
> *lgp, long *timeout, gfp_t gfp_flags)
> &lgp->res.stateid,
> status);
>
> + nfs4_sequence_free_slot(&lgp->res.seq_res);
> /* if layoutp->len is 0, nfs4_layoutget_prepare called
> rpc_exit */
> if (status == 0 && lgp->res.layoutp->len)
> lseg = pnfs_layout_process(lgp);
> - nfs4_sequence_free_slot(&lgp->res.seq_res);
> rpc_put_task(task);
> dprintk("<-- %s status=%d\n", __func__, status);
> if (status)
The placement of the call to nfs4_sequence_free_slot() after the call
to pnfs_layout_process() is deliberate, and is needed in order to
ensure that layoutrecall callbacks are processed correctly in order
after the layoutget has been processed (see referring_call_exists()).
The call to nfs4_find_get_deviceid() needs to be moved out of
pnfs_update_layout().
In the flexfiles driver, it was moved to nfs4_ff_layout_prepare_ds(),
but it looks to me as if the files driver needs to call it earlier due
to the need to calculate stripes. Perhaps the right thing is to add a
wrapper around pnfs_update_layout() in fs/nfs/filelayout/filelayout.c
that can be called by both filelayout_pg_init_read() and
filelayout_pg_init_write()?
--
Trond Myklebust
Linux NFS client maintainer, PrimaryData
trond.myklebust@xxxxxxxxxxxxxxx
��.n��������+%������w��{.n�����{��w��� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
