On Mon, 2012-02-13 at 17:22 -0500, Weston Andros Adamson wrote:
> diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
> index efe5495..ab7a28f 100644
> --- a/net/sunrpc/xprt.c
> +++ b/net/sunrpc/xprt.c
> @@ -1132,7 +1132,10 @@ void xprt_release(struct rpc_task *task)
> return;
>
> xprt = req->rq_xprt;
> - rpc_count_iostats(task);
> + if (task->tk_client)
> + rpc_count_iostats(task, task->tk_client->cl_metrics);
> + if (task->tk_ops->rpc_count_stats != NULL)
> + task->tk_ops->rpc_count_stats(task, task->tk_calldata);
> spin_lock_bh(&xprt->transport_lock);
> xprt->ops->release_xprt(xprt, task);
> if (xprt->ops->release_request)
2 nits:
1. Aren't we guaranteed that task->tk_client != NULL in
xprt_release()?
2. Shouldn't we be calling _either_ rpc_count_iostats(), _or_
task->tk_ops->rpc_count_stats()? As far as I can see, the
nfsstat output will now be double-counting and pNFS reads,
writes and commits that are sent to the DS.
Cheers
Trond
--
Trond Myklebust
Linux NFS client maintainer
NetApp
Trond.Myklebust@xxxxxxxxxx
www.netapp.com
��.n��������+%������w��{.n�����{��w��� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
