svcrdma: Convert rdma_stat_sq_starve to a per-CPU counter

Avoid the overhead of a memory bus lock cycle for counting a value
that is hardly every used.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

include/linux/sunrpc/svc_rdma.h

@@ -69,7 +69,7 @@ extern unsigned int svcrdma_max_req_size;
 extern struct percpu_counter svcrdma_stat_recv;
 extern atomic_t rdma_stat_read;
 extern atomic_t rdma_stat_write;
-extern atomic_t rdma_stat_sq_starve;
+extern struct percpu_counter svcrdma_stat_sq_starve;
 extern atomic_t rdma_stat_rq_starve;
 extern atomic_t rdma_stat_rq_poll;
 extern atomic_t rdma_stat_rq_prod;

net/sunrpc/xprtrdma/svc_rdma.c

@@ -66,7 +66,7 @@ static unsigned int max_max_inline = RPCRDMA_MAX_INLINE_THRESH;
 struct percpu_counter svcrdma_stat_recv;
 atomic_t rdma_stat_read;
 atomic_t rdma_stat_write;
-atomic_t rdma_stat_sq_starve;
+struct percpu_counter svcrdma_stat_sq_starve;
 atomic_t rdma_stat_rq_starve;
 atomic_t rdma_stat_rq_poll;
 atomic_t rdma_stat_rq_prod;
@@ -199,10 +199,10 @@ static struct ctl_table svcrdma_parm_table[] = {
 	},
 	{
 		.procname	= "rdma_stat_sq_starve",
-		.data		= &rdma_stat_sq_starve,
-		.maxlen		= sizeof(atomic_t),
+		.data		= &svcrdma_stat_sq_starve,
+		.maxlen		= SVCRDMA_COUNTER_BUFSIZ,
 		.mode		= 0644,
-		.proc_handler	= read_reset_stat,
+		.proc_handler	= svcrdma_counter_handler,
 	},
 	{
 		.procname	= "rdma_stat_rq_starve",
@@ -267,6 +267,7 @@ static void svc_rdma_proc_cleanup(void)
 	unregister_sysctl_table(svcrdma_table_header);
 	svcrdma_table_header = NULL;
 
+	percpu_counter_destroy(&svcrdma_stat_sq_starve);
 	percpu_counter_destroy(&svcrdma_stat_recv);
 }
 
@@ -278,6 +279,9 @@ static int svc_rdma_proc_init(void)
 		return 0;
 
 	rc = percpu_counter_init(&svcrdma_stat_recv, 0, GFP_KERNEL);
+	if (rc)
+		goto out_err;
+	rc = percpu_counter_init(&svcrdma_stat_sq_starve, 0, GFP_KERNEL);
 	if (rc)
 		goto out_err;
 
@@ -285,6 +289,7 @@ static int svc_rdma_proc_init(void)
 	return 0;
 
 out_err:
+	percpu_counter_destroy(&svcrdma_stat_recv);
 	return rc;
 }
 

net/sunrpc/xprtrdma/svc_rdma_rw.c

@@ -364,6 +364,7 @@ static int svc_rdma_post_chunk_ctxt(struct svc_rdma_chunk_ctxt *cc)
 			return 0;
 		}
 
+		percpu_counter_inc(&svcrdma_stat_sq_starve);
 		trace_svcrdma_sq_full(rdma);
 		atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
 		wait_event(rdma->sc_send_wait,

net/sunrpc/xprtrdma/svc_rdma_sendto.c

@@ -317,7 +317,7 @@ int svc_rdma_send(struct svcxprt_rdma *rdma, struct svc_rdma_send_ctxt *ctxt)
 	/* If the SQ is full, wait until an SQ entry is available */
 	while (1) {
 		if ((atomic_dec_return(&rdma->sc_sq_avail) < 0)) {
-			atomic_inc(&rdma_stat_sq_starve);
+			percpu_counter_inc(&svcrdma_stat_sq_starve);
 			trace_svcrdma_sq_full(rdma);
 			atomic_inc(&rdma->sc_sq_avail);
 			wait_event(rdma->sc_send_wait,