[PATCH] RPC: add generic interface for adjusting the congestion window

 A new interface that allows transports to adjust their congestion window
 using the Van Jacobson implementation in xprt.c is provided.

 Test-plan:
 Use WAN simulation to cause sporadic bursty packet loss.  Look for
 significant regression in performance or client stability.

 Signed-off-by: Chuck Lever <cel@netapp.com>
 Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

3 files changed, 33 insertions, 49 deletions

diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
index 443c3f984cf9..2e48752d55d9 100644
--- a/include/linux/sunrpc/xprt.h
+++ b/include/linux/sunrpc/xprt.h
@@ -258,9 +258,10 @@ void			xprt_set_retrans_timeout_rtt(struct rpc_task *task);
 void                    xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status);
 void                    xprt_wait_for_buffer_space(struct rpc_task *task);
 void                    xprt_write_space(struct rpc_xprt *xprt);
+void                    xprt_update_rtt(struct rpc_task *task);
 void                    xprt_adjust_cwnd(struct rpc_task *task, int result);
 struct rpc_rqst *       xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid);
-void                    xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied);
+void                    xprt_complete_rqst(struct rpc_task *task, int copied);
 void                    xprt_disconnect(struct rpc_xprt *xprt);
 
 /*
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index ffc595592af3..707806fe1a23 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -592,55 +592,41 @@ struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
 }
 
 /**
+ * xprt_update_rtt - update an RPC client's RTT state after receiving a reply
+ * @task: RPC request that recently completed
+ *
+ */
+void xprt_update_rtt(struct rpc_task *task)
+{
+        struct rpc_rqst *req = task->tk_rqstp;
+        struct rpc_rtt *rtt = task->tk_client->cl_rtt;
+        unsigned timer = task->tk_msg.rpc_proc->p_timer;
+
+        if (timer) {
+                if (req->rq_ntrans == 1)
+                        rpc_update_rtt(rtt, timer,
+                                        (long)jiffies - req->rq_xtime);
+                rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
+        }
+}
+
+/**
  * xprt_complete_rqst - called when reply processing is complete
- * @xprt: controlling transport
- * @req: RPC request that just completed
+ * @task: RPC request that recently completed
  * @copied: actual number of bytes received from the transport
  *
+ * Caller holds transport lock.
  */
-void xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
-{
-        struct rpc_task *task = req->rq_task;
-        struct rpc_clnt *clnt = task->tk_client;
-
-        /* Adjust congestion window */
-        if (!xprt->nocong) {
-                unsigned timer = task->tk_msg.rpc_proc->p_timer;
-                xprt_adjust_cwnd(task, copied);
-                if (timer) {
-                        if (req->rq_ntrans == 1)
-                                rpc_update_rtt(clnt->cl_rtt, timer,
-                                                (long)jiffies - req->rq_xtime);
-                        rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1);
-                }
-        }
+void xprt_complete_rqst(struct rpc_task *task, int copied)
+{
+        struct rpc_rqst *req = task->tk_rqstp;
 
-#ifdef RPC_PROFILE
-        /* Profile only reads for now */
-        if (copied > 1024) {
-                static unsigned long    nextstat;
-                static unsigned long    pkt_rtt, pkt_len, pkt_cnt;
-
-                pkt_cnt++;
-                pkt_len += req->rq_slen + copied;
-                pkt_rtt += jiffies - req->rq_xtime;
-                if (time_before(nextstat, jiffies)) {
-                        printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
-                        printk("RPC: %ld %ld %ld %ld stat\n",
-                                        jiffies, pkt_cnt, pkt_len, pkt_rtt);
-                        pkt_rtt = pkt_len = pkt_cnt = 0;
-                        nextstat = jiffies + 5 * HZ;
-                }
-        }
-#endif
+        dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
+                        task->tk_pid, ntohl(req->rq_xid), copied);
 
-        dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
         list_del_init(&req->rq_list);
         req->rq_received = req->rq_private_buf.len = copied;
-
-        /* ... and wake up the process. */
         rpc_wake_up_task(task);
-        return;
 }
 
 static void xprt_timer(struct rpc_task *task)
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index c3658ff027a6..980f26504f48 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -460,8 +460,6 @@ static void xs_udp_data_ready(struct sock *sk, int len)
                 goto out_unlock;
         task = rovr->rq_task;
 
-        dprintk("RPC: %4d received reply\n", task->tk_pid);
-
         if ((copied = rovr->rq_private_buf.buflen) > repsize)
                 copied = repsize;
 
@@ -472,7 +470,9 @@ static void xs_udp_data_ready(struct sock *sk, int len)
         /* Something worked... */
         dst_confirm(skb->dst);
 
-        xprt_complete_rqst(xprt, rovr, copied);
+        xprt_adjust_cwnd(task, copied);
+        xprt_update_rtt(task);
+        xprt_complete_rqst(task, copied);
 
  out_unlock:
         spin_unlock(&xprt->transport_lock);
@@ -634,11 +634,8 @@ static inline void xs_tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc
         }
 
 out:
-        if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
-                dprintk("RPC: %4d received reply complete\n",
-                                req->rq_task->tk_pid);
-                xprt_complete_rqst(xprt, req, xprt->tcp_copied);
-        }
+        if (!(xprt->tcp_flags & XPRT_COPY_DATA))
+                xprt_complete_rqst(req->rq_task, xprt->tcp_copied);
         spin_unlock(&xprt->transport_lock);
         xs_tcp_check_recm(xprt);
 }