SUNRPC: Make RTT measurement more precise (Receive)

Some RPC transports have more overhead in their reply handlers
than others. For example, for RPC-over-RDMA:

- RPC completion has to wait for memory invalidation, which is
  not a part of the server/network round trip

- Recently a context switch was introduced into the reply handler,
  which further artificially inflates the measure of RPC RTT

To capture just server and network latencies more precisely: when
receiving a reply, compute the RTT as soon as the XID is recognized
rather than at RPC completion time.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

1 files changed, 3 insertions, 2 deletions

diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 13fbb4849188..cb7784c5b741 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -826,6 +826,7 @@ static void xprt_connect_status(struct rpc_task *task)
  * @xprt: transport on which the original request was transmitted
  * @xid: RPC XID of incoming reply
  *
+ * Caller holds xprt->recv_lock.
  */
 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
 {
@@ -834,6 +835,7 @@ struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
 	list_for_each_entry(entry, &xprt->recv, rq_list)
 		if (entry->rq_xid == xid) {
 			trace_xprt_lookup_rqst(xprt, xid, 0);
+			entry->rq_rtt = ktime_sub(ktime_get(), entry->rq_xtime);
 			return entry;
 		}
 
@@ -915,7 +917,7 @@ EXPORT_SYMBOL_GPL(xprt_update_rtt);
  * @task: RPC request that recently completed
  * @copied: actual number of bytes received from the transport
  *
- * Caller holds transport lock.
+ * Caller holds xprt->recv_lock.
  */
 void xprt_complete_rqst(struct rpc_task *task, int copied)
 {
@@ -927,7 +929,6 @@ void xprt_complete_rqst(struct rpc_task *task, int copied)
 	trace_xprt_complete_rqst(xprt, req->rq_xid, copied);
 
 	xprt->stat.recvs++;
-	req->rq_rtt = ktime_sub(ktime_get(), req->rq_xtime);
 
 	list_del_init(&req->rq_list);
 	req->rq_private_buf.len = copied;