From e456c835369f0d06134fbb3ca516f931d2cd5c5b Mon Sep 17 00:00:00 2001
From: Daniel Hodges <hodges.daniel.scott@gmail.com>
Date: Fri, 11 Oct 2024 20:27:44 -0700
Subject: [PATCH] scx_layered: Add topology integration test

Add a bpftrace script that does a topology aware test. The test script
runs a bpftrace script that asserts that stress-ng processes are
scheduled on NUMA node 0 only.

Signed-off-by: Daniel Hodges <hodges.daniel.scott@gmail.com>
---
 .../scx_layered/integration/layer_node.bt     | 48 +++++++++++++++++++
 scheds/rust/scx_layered/integration/numa.json | 43 +++++++++++++++++
 .../scx_layered/integration/numa_tests.sh     | 32 +++++++++++++
 3 files changed, 123 insertions(+)
 create mode 100755 scheds/rust/scx_layered/integration/layer_node.bt
 create mode 100644 scheds/rust/scx_layered/integration/numa.json
 create mode 100755 scheds/rust/scx_layered/integration/numa_tests.sh

diff --git a/scheds/rust/scx_layered/integration/layer_node.bt b/scheds/rust/scx_layered/integration/layer_node.bt
new file mode 100755
index 0000000..662ee9b
--- /dev/null
+++ b/scheds/rust/scx_layered/integration/layer_node.bt
@@ -0,0 +1,48 @@
+#!/usr/bin/env -S bpftrace --unsafe -q
+
+/*
+ * Asserts that the `node` layer config works properly by failing if the pid
+ * passed to the script runs on NUMA node 1. The layered config should restrict
+ * the pid passed to the script to run on a layer that only runs on NUMA node 0.
+ */
+
+BEGIN
+{
+  @bpftrace_pid = pid;
+  @sig = 0;
+
+  if ($1 == 0) {
+	// exit 137
+	@sig = 9;
+  }
+}
+
+profile:hz:1
+{
+    @counts[cpu] = @counts[cpu] + 1;
+    if (@counts[cpu] == 15) {
+	// exit 0
+	@sig = 15;
+    }
+}
+
+rawtracepoint:sched_switch
+{
+	$task = (struct task_struct *)arg1;
+
+	if (($task->parent->pid == $1 && numaid == 1) ||
+	    ($task->real_parent->pid == $1 && numaid == 1)) {
+		// exit 137
+		@sig = 9;
+	}
+}
+
+kprobe:__x64_sys_* / @bpftrace_pid == pid / {
+	if (@sig > 0) {
+		signal(@sig);
+	}
+}
+
+interval:s:1 {
+	print(("bpftrace monitoring pid", $1, "signal", @sig));
+}
diff --git a/scheds/rust/scx_layered/integration/numa.json b/scheds/rust/scx_layered/integration/numa.json
new file mode 100644
index 0000000..1ce161e
--- /dev/null
+++ b/scheds/rust/scx_layered/integration/numa.json
@@ -0,0 +1,43 @@
+[
+  {
+    "name": "numa_0",
+    "comment": "numa0",
+    "matches": [
+      [
+        {
+          "CommPrefix": "stress-ng"
+        }
+      ],
+      [
+        {
+          "PcommPrefix": "stress-ng"
+        }
+      ]
+    ],
+    "kind": {
+      "Confined": {
+        "util_range": [
+          0.4,
+          0.90
+        ],
+	"growth_algo": "Linear",
+        "nodes": [
+          0
+        ]
+      }
+    }
+  },
+  {
+    "name": "reset",
+    "comment": "the rest",
+    "matches":[[]],
+    "kind": {
+      "Grouped": {
+        "util_range": [
+          0.05,
+          0.60
+        ]
+      }
+    }
+  }
+]
diff --git a/scheds/rust/scx_layered/integration/numa_tests.sh b/scheds/rust/scx_layered/integration/numa_tests.sh
new file mode 100755
index 0000000..9e8b260
--- /dev/null
+++ b/scheds/rust/scx_layered/integration/numa_tests.sh
@@ -0,0 +1,32 @@
+#!/usr/bin/env bash
+
+layered_bin=${1:-../../../../target/release/scx_layered}
+test_scripts=( "layer_node.bt" )
+
+for test_script in "${test_scripts[@]}"; do
+	sudo pkill -9 -f scx_layered
+	sudo "${layered_bin}" --stats 1 f:numa.json -v &
+	layered_pid=$!
+
+	echo "layered pid ${layered_pid}"
+	sleep 2
+
+	stress-ng -c 2 -f 1 -t 30 &
+	stress_pid=$!
+
+	echo "stress-ng pid ${stress_pid}"
+	sleep 1
+
+	sudo "./${test_script}" "${stress_pid}"
+	test_exit=$?
+
+	pidof scx_layered && sudo pkill -9 -f scx_layered
+	# always cleanup stress-ng
+	sudo pkill -9 -f stress-ng
+
+	if [ $test_exit -ne 0 ]; then
+		echo "test script ${test_script} failed: ${test_exit}"
+		exit $test_exit;
+	fi
+	echo "test script ${test_script} passed: ${test_exit}"
+done