scftorture: Count single_rpc offline failures in statistics outputgrafted

scf_torture_stats_print() aggregates each invoker thread's counters
into a local scf_statistics structure before printing, but the
n_single_rpc_ofl field is missing from the aggregation loop.  As a
result, the "single_rpc_ofl" value printed in the statistics line is
always zero, even when smp_call_function_single() invocations for the
RPC test have failed due to offline CPUs and been counted by the
invoker threads.

Add the missing accumulation so that the printed value reflects the
actual counts.

Signed-off-by: Joel Fernandes <joelagnelf@nvidia.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>

1 files changed, 103 insertions, 0 deletions

diff --git a/tools/perf/util/affinity.c b/tools/perf/util/affinity.c
new file mode 100644
index 000000000..6c64b5f69
--- /dev/null
+++ b/tools/perf/util/affinity.c
@@ -0,0 +1,103 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Manage affinity to optimize IPIs inside the kernel perf API. */
+#define _GNU_SOURCE 1
+#include <sched.h>
+#include <stdlib.h>
+#include <linux/bitmap.h>
+#include <linux/zalloc.h>
+#include <perf/cpumap.h>
+#include "perf.h"
+#include "cpumap.h"
+#include "affinity.h"
+
+static int get_cpu_set_size(void)
+{
+	int sz = cpu__max_cpu().cpu + 8 - 1;
+	/*
+	 * sched_getaffinity doesn't like masks smaller than the kernel.
+	 * Hopefully that's big enough.
+	 */
+	if (sz < 4096)
+		sz = 4096;
+	return sz / 8;
+}
+
+int affinity__setup(struct affinity *a)
+{
+	int cpu_set_size = get_cpu_set_size();
+
+	a->orig_cpus = bitmap_zalloc(cpu_set_size * 8);
+	if (!a->orig_cpus)
+		return -1;
+	sched_getaffinity(0, cpu_set_size, (cpu_set_t *)a->orig_cpus);
+	a->sched_cpus = bitmap_zalloc(cpu_set_size * 8);
+	if (!a->sched_cpus) {
+		zfree(&a->orig_cpus);
+		return -1;
+	}
+	bitmap_zero((unsigned long *)a->sched_cpus, cpu_set_size);
+	a->changed = false;
+	return 0;
+}
+
+/*
+ * perf_event_open does an IPI internally to the target CPU.
+ * It is more efficient to change perf's affinity to the target
+ * CPU and then set up all events on that CPU, so we amortize
+ * CPU communication.
+ */
+void affinity__set(struct affinity *a, int cpu)
+{
+	int cpu_set_size = get_cpu_set_size();
+
+	/*
+	 * Return:
+	 * - if cpu is -1
+	 * - restrict out of bound access to sched_cpus
+	 */
+	if (cpu == -1 || ((cpu >= (cpu_set_size * 8))))
+		return;
+
+	a->changed = true;
+	__set_bit(cpu, a->sched_cpus);
+	/*
+	 * We ignore errors because affinity is just an optimization.
+	 * This could happen for example with isolated CPUs or cpusets.
+	 * In this case the IPIs inside the kernel's perf API still work.
+	 */
+	sched_setaffinity(0, cpu_set_size, (cpu_set_t *)a->sched_cpus);
+	__clear_bit(cpu, a->sched_cpus);
+}
+
+static void __affinity__cleanup(struct affinity *a)
+{
+	int cpu_set_size = get_cpu_set_size();
+
+	if (a->changed)
+		sched_setaffinity(0, cpu_set_size, (cpu_set_t *)a->orig_cpus);
+	zfree(&a->sched_cpus);
+	zfree(&a->orig_cpus);
+}
+
+void affinity__cleanup(struct affinity *a)
+{
+	if (a != NULL)
+		__affinity__cleanup(a);
+}
+
+void cpu_map__set_affinity(const struct perf_cpu_map *cpumap)
+{
+	int cpu_set_size = get_cpu_set_size();
+	unsigned long *cpuset = bitmap_zalloc(cpu_set_size * 8);
+	struct perf_cpu cpu;
+	unsigned int idx;
+
+	if (!cpuset)
+		return;
+
+	perf_cpu_map__for_each_cpu_skip_any(cpu, idx, cpumap)
+		__set_bit(cpu.cpu, cpuset);
+
+	sched_setaffinity(0, cpu_set_size, (cpu_set_t *)cpuset);
+	zfree(&cpuset);
+}