SUNRPC: Use per-CPU counters to tally server RPC counts

 - Improves counting accuracy
 - Reduces cross-CPU memory traffic
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

fs/lockd/svc.c

@@ -721,7 +721,7 @@ static int nlmsvc_dispatch(struct svc_rqst *rqstp)
 /*
  * Define NLM program and procedures
  */
-static unsigned int nlmsvc_version1_count[17];
+static DEFINE_PER_CPU_ALIGNED(unsigned long, nlmsvc_version1_count[17]);
 static const struct svc_version	nlmsvc_version1 = {
 	.vs_vers	= 1,
 	.vs_nproc	= 17,
@@ -730,26 +730,31 @@ static const struct svc_version	nlmsvc_version1 = {
 	.vs_dispatch	= nlmsvc_dispatch,
 	.vs_xdrsize	= NLMSVC_XDRSIZE,
 };
-static unsigned int nlmsvc_version3_count[24];
+
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nlmsvc_version3_count[ARRAY_SIZE(nlmsvc_procedures)]);
 static const struct svc_version	nlmsvc_version3 = {
 	.vs_vers	= 3,
-	.vs_nproc	= 24,
+	.vs_nproc	= ARRAY_SIZE(nlmsvc_procedures),
 	.vs_proc	= nlmsvc_procedures,
 	.vs_count	= nlmsvc_version3_count,
 	.vs_dispatch	= nlmsvc_dispatch,
 	.vs_xdrsize	= NLMSVC_XDRSIZE,
 };
+
 #ifdef CONFIG_LOCKD_V4
-static unsigned int nlmsvc_version4_count[24];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nlmsvc_version4_count[ARRAY_SIZE(nlmsvc_procedures4)]);
 static const struct svc_version	nlmsvc_version4 = {
 	.vs_vers	= 4,
-	.vs_nproc	= 24,
+	.vs_nproc	= ARRAY_SIZE(nlmsvc_procedures4),
 	.vs_proc	= nlmsvc_procedures4,
 	.vs_count	= nlmsvc_version4_count,
 	.vs_dispatch	= nlmsvc_dispatch,
 	.vs_xdrsize	= NLMSVC_XDRSIZE,
 };
 #endif
+
 static const struct svc_version *nlmsvc_version[] = {
 	[1] = &nlmsvc_version1,
 	[3] = &nlmsvc_version3,

fs/nfs/callback_xdr.c

@@ -1069,7 +1069,8 @@ static const struct svc_procedure nfs4_callback_procedures1[] = {
 	}
 };
 
-static unsigned int nfs4_callback_count1[ARRAY_SIZE(nfs4_callback_procedures1)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfs4_callback_count1[ARRAY_SIZE(nfs4_callback_procedures1)]);
 const struct svc_version nfs4_callback_version1 = {
 	.vs_vers = 1,
 	.vs_nproc = ARRAY_SIZE(nfs4_callback_procedures1),
@@ -1081,7 +1082,8 @@ const struct svc_version nfs4_callback_version1 = {
 	.vs_need_cong_ctrl = true,
 };
 
-static unsigned int nfs4_callback_count4[ARRAY_SIZE(nfs4_callback_procedures1)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfs4_callback_count4[ARRAY_SIZE(nfs4_callback_procedures1)]);
 const struct svc_version nfs4_callback_version4 = {
 	.vs_vers = 4,
 	.vs_nproc = ARRAY_SIZE(nfs4_callback_procedures1),

fs/nfsd/nfs2acl.c

@@ -377,10 +377,11 @@ static const struct svc_procedure nfsd_acl_procedures2[5] = {
 	},
 };
 
-static unsigned int nfsd_acl_count2[ARRAY_SIZE(nfsd_acl_procedures2)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfsd_acl_count2[ARRAY_SIZE(nfsd_acl_procedures2)]);
 const struct svc_version nfsd_acl_version2 = {
 	.vs_vers	= 2,
-	.vs_nproc	= 5,
+	.vs_nproc	= ARRAY_SIZE(nfsd_acl_procedures2),
 	.vs_proc	= nfsd_acl_procedures2,
 	.vs_count	= nfsd_acl_count2,
 	.vs_dispatch	= nfsd_dispatch,

fs/nfsd/nfs3acl.c

@@ -266,10 +266,11 @@ static const struct svc_procedure nfsd_acl_procedures3[3] = {
 	},
 };
 
-static unsigned int nfsd_acl_count3[ARRAY_SIZE(nfsd_acl_procedures3)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfsd_acl_count3[ARRAY_SIZE(nfsd_acl_procedures3)]);
 const struct svc_version nfsd_acl_version3 = {
 	.vs_vers	= 3,
-	.vs_nproc	= 3,
+	.vs_nproc	= ARRAY_SIZE(nfsd_acl_procedures3),
 	.vs_proc	= nfsd_acl_procedures3,
 	.vs_count	= nfsd_acl_count3,
 	.vs_dispatch	= nfsd_dispatch,

fs/nfsd/nfs3proc.c

@@ -1064,10 +1064,11 @@ static const struct svc_procedure nfsd_procedures3[22] = {
 	},
 };
 
-static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures3)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfsd_count3[ARRAY_SIZE(nfsd_procedures3)]);
 const struct svc_version nfsd_version3 = {
 	.vs_vers	= 3,
-	.vs_nproc	= 22,
+	.vs_nproc	= ARRAY_SIZE(nfsd_procedures3),
 	.vs_proc	= nfsd_procedures3,
 	.vs_dispatch	= nfsd_dispatch,
 	.vs_count	= nfsd_count3,

fs/nfsd/nfs4proc.c

@@ -3596,12 +3596,13 @@ static const struct svc_procedure nfsd_procedures4[2] = {
 	},
 };
 
-static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
 const struct svc_version nfsd_version4 = {
 	.vs_vers		= 4,
-	.vs_nproc		= 2,
+	.vs_nproc		= ARRAY_SIZE(nfsd_procedures4),
 	.vs_proc		= nfsd_procedures4,
-	.vs_count		= nfsd_count3,
+	.vs_count		= nfsd_count4,
 	.vs_dispatch		= nfsd_dispatch,
 	.vs_xdrsize		= NFS4_SVC_XDRSIZE,
 	.vs_rpcb_optnl		= true,

fs/nfsd/nfsproc.c

@@ -838,11 +838,11 @@ static const struct svc_procedure nfsd_procedures2[18] = {
 	},
 };
 
-
-static unsigned int nfsd_count2[ARRAY_SIZE(nfsd_procedures2)];
+static DEFINE_PER_CPU_ALIGNED(unsigned long,
+			      nfsd_count2[ARRAY_SIZE(nfsd_procedures2)]);
 const struct svc_version nfsd_version2 = {
 	.vs_vers	= 2,
-	.vs_nproc	= 18,
+	.vs_nproc	= ARRAY_SIZE(nfsd_procedures2),
 	.vs_proc	= nfsd_procedures2,
 	.vs_count	= nfsd_count2,
 	.vs_dispatch	= nfsd_dispatch,

include/linux/lockd/lockd.h

@@ -196,9 +196,9 @@ struct nlm_block {
  * Global variables
  */
 extern const struct rpc_program	nlm_program;
-extern const struct svc_procedure nlmsvc_procedures[];
+extern const struct svc_procedure nlmsvc_procedures[24];
 #ifdef CONFIG_LOCKD_V4
-extern const struct svc_procedure nlmsvc_procedures4[];
+extern const struct svc_procedure nlmsvc_procedures4[24];
 #endif
 extern int			nlmsvc_grace_period;
 extern unsigned long		nlmsvc_timeout;

include/linux/sunrpc/svc.h

@@ -377,7 +377,7 @@ struct svc_version {
 	u32			vs_vers;	/* version number */
 	u32			vs_nproc;	/* number of procedures */
 	const struct svc_procedure *vs_proc;	/* per-procedure info */
-	unsigned int		*vs_count;	/* call counts */
+	unsigned long __percpu	*vs_count;	/* call counts */
 	u32			vs_xdrsize;	/* xdrsize needed for this version */
 
 	/* Don't register with rpcbind */

net/sunrpc/stats.c

@@ -83,7 +83,8 @@ void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp)
 {
 	const struct svc_program *prog = statp->program;
 	const struct svc_version *vers;
-	unsigned int i, j;
+	unsigned int i, j, k;
+	unsigned long count;
 
 	seq_printf(seq,
 		"net %u %u %u %u\n",
@@ -104,8 +105,12 @@ void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp)
 		if (!vers)
 			continue;
 		seq_printf(seq, "proc%d %u", i, vers->vs_nproc);
-		for (j = 0; j < vers->vs_nproc; j++)
-			seq_printf(seq, " %u", vers->vs_count[j]);
+		for (j = 0; j < vers->vs_nproc; j++) {
+			count = 0;
+			for_each_possible_cpu(k)
+				count += per_cpu(vers->vs_count[j], k);
+			seq_printf(seq, " %lu", count);
+		}
 		seq_putc(seq, '\n');
 	}
 }

net/sunrpc/svc.c

@@ -1208,7 +1208,7 @@ svc_generic_init_request(struct svc_rqst *rqstp,
 	memset(rqstp->rq_resp, 0, procp->pc_ressize);
 
 	/* Bump per-procedure stats counter */
-	versp->vs_count[rqstp->rq_proc]++;
+	this_cpu_inc(versp->vs_count[rqstp->rq_proc]);
 
 	ret->dispatch = versp->vs_dispatch;
 	return rpc_success;