Merge tag 'sched_ext-for-7.0-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext

Pull sched_ext fixes from Tejun Heo:

 - Various bug fixes for the example schedulers and selftests

* tag 'sched_ext-for-7.0-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext:
  tools/sched_ext: fix getopt not re-parsed on restart
  tools/sched_ext: scx_userland: fix data races on shared counters
  tools/sched_ext: scx_pair: fix stride == 0 crash on single-CPU systems
  tools/sched_ext: scx_central: fix CPU_SET and skeleton leak on early exit
  tools/sched_ext: scx_userland: fix stale data on restart
  tools/sched_ext: scx_flatcg: fix potential stack overflow from VLA in fcg_read_stats
  selftests/sched_ext: Fix rt_stall flaky failure
  tools/sched_ext: scx_userland: fix restart and stats thread lifecycle bugs
  tools/sched_ext: scx_central: fix sched_setaffinity() call with the set size
  tools/sched_ext: scx_flatcg: zero-initialize stats counter array

8 files changed, 96 insertions, 18 deletions

diff --git a/tools/sched_ext/scx_central.c b/tools/sched_ext/scx_central.c
index 55931a4cd71c..2a805f1d6c8f 100644
--- a/tools/sched_ext/scx_central.c
+++ b/tools/sched_ext/scx_central.c
@@ -50,11 +50,13 @@ int main(int argc, char **argv)
 	__u64 seq = 0, ecode;
 	__s32 opt;
 	cpu_set_t *cpuset;
+	size_t cpuset_size;
 
 	libbpf_set_print(libbpf_print_fn);
 	signal(SIGINT, sigint_handler);
 	signal(SIGTERM, sigint_handler);
 restart:
+	optind = 1;
 	skel = SCX_OPS_OPEN(central_ops, scx_central);
 
 	skel->rodata->central_cpu = 0;
@@ -73,6 +75,7 @@ restart:
 			u32 central_cpu = strtoul(optarg, NULL, 0);
 			if (central_cpu >= skel->rodata->nr_cpu_ids) {
 				fprintf(stderr, "invalid central CPU id value, %u given (%u max)\n", central_cpu, skel->rodata->nr_cpu_ids);
+				scx_central__destroy(skel);
 				return -1;
 			}
 			skel->rodata->central_cpu = (s32)central_cpu;
@@ -106,9 +109,10 @@ restart:
 	 */
 	cpuset = CPU_ALLOC(skel->rodata->nr_cpu_ids);
 	SCX_BUG_ON(!cpuset, "Failed to allocate cpuset");
-	CPU_ZERO_S(CPU_ALLOC_SIZE(skel->rodata->nr_cpu_ids), cpuset);
-	CPU_SET(skel->rodata->central_cpu, cpuset);
-	SCX_BUG_ON(sched_setaffinity(0, sizeof(*cpuset), cpuset),
+	cpuset_size = CPU_ALLOC_SIZE(skel->rodata->nr_cpu_ids);
+	CPU_ZERO_S(cpuset_size, cpuset);
+	CPU_SET_S(skel->rodata->central_cpu, cpuset_size, cpuset);
+	SCX_BUG_ON(sched_setaffinity(0, cpuset_size, cpuset),
 		   "Failed to affinitize to central CPU %d (max %d)",
 		   skel->rodata->central_cpu, skel->rodata->nr_cpu_ids - 1);
 	CPU_FREE(cpuset);
diff --git a/tools/sched_ext/scx_cpu0.c b/tools/sched_ext/scx_cpu0.c
index 1e4fa4ab8da9..a6fba9978b9c 100644
--- a/tools/sched_ext/scx_cpu0.c
+++ b/tools/sched_ext/scx_cpu0.c
@@ -69,6 +69,7 @@ int main(int argc, char **argv)
 	signal(SIGINT, sigint_handler);
 	signal(SIGTERM, sigint_handler);
 restart:
+	optind = 1;
 	skel = SCX_OPS_OPEN(cpu0_ops, scx_cpu0);
 
 	skel->rodata->nr_cpus = libbpf_num_possible_cpus();
diff --git a/tools/sched_ext/scx_flatcg.c b/tools/sched_ext/scx_flatcg.c
index cd85eb401179..d865c381589b 100644
--- a/tools/sched_ext/scx_flatcg.c
+++ b/tools/sched_ext/scx_flatcg.c
@@ -102,21 +102,27 @@ static float read_cpu_util(__u64 *last_sum, __u64 *last_idle)
 
 static void fcg_read_stats(struct scx_flatcg *skel, __u64 *stats)
 {
-	__u64 cnts[FCG_NR_STATS][skel->rodata->nr_cpus];
+	__u64 *cnts;
 	__u32 idx;
 
+	cnts = calloc(skel->rodata->nr_cpus, sizeof(__u64));
+	if (!cnts)
+		return;
+
 	memset(stats, 0, sizeof(stats[0]) * FCG_NR_STATS);
 
 	for (idx = 0; idx < FCG_NR_STATS; idx++) {
 		int ret, cpu;
 
 		ret = bpf_map_lookup_elem(bpf_map__fd(skel->maps.stats),
-					  &idx, cnts[idx]);
+					  &idx, cnts);
 		if (ret < 0)
 			continue;
 		for (cpu = 0; cpu < skel->rodata->nr_cpus; cpu++)
-			stats[idx] += cnts[idx][cpu];
+			stats[idx] += cnts[cpu];
 	}
+
+	free(cnts);
 }
 
 int main(int argc, char **argv)
@@ -135,6 +141,7 @@ int main(int argc, char **argv)
 	signal(SIGINT, sigint_handler);
 	signal(SIGTERM, sigint_handler);
 restart:
+	optind = 1;
 	skel = SCX_OPS_OPEN(flatcg_ops, scx_flatcg);
 
 	skel->rodata->nr_cpus = libbpf_num_possible_cpus();
diff --git a/tools/sched_ext/scx_pair.c b/tools/sched_ext/scx_pair.c
index d3e97faa6334..2e509391f3da 100644
--- a/tools/sched_ext/scx_pair.c
+++ b/tools/sched_ext/scx_pair.c
@@ -53,10 +53,10 @@ int main(int argc, char **argv)
 	signal(SIGINT, sigint_handler);
 	signal(SIGTERM, sigint_handler);
 restart:
+	optind = 1;
 	skel = SCX_OPS_OPEN(pair_ops, scx_pair);
 
 	skel->rodata->nr_cpu_ids = libbpf_num_possible_cpus();
-	assert(skel->rodata->nr_cpu_ids > 0);
 	skel->rodata->pair_batch_dur_ns = __COMPAT_ENUM_OR_ZERO("scx_public_consts", "SCX_SLICE_DFL");
 
 	/* pair up the earlier half to the latter by default, override with -s */
@@ -76,6 +76,12 @@ restart:
 		}
 	}
 
+	/* Stride must be positive to pair distinct CPUs. */
+	if (stride <= 0) {
+		fprintf(stderr, "Invalid stride %d, must be positive\n", stride);
+		scx_pair__destroy(skel);
+		return -1;
+	}
 	bpf_map__set_max_entries(skel->maps.pair_ctx, skel->rodata->nr_cpu_ids / 2);
 
 	/* Resize arrays so their element count is equal to cpu count. */
diff --git a/tools/sched_ext/scx_sdt.c b/tools/sched_ext/scx_sdt.c
index b0363363476d..d8ca9aa316a5 100644
--- a/tools/sched_ext/scx_sdt.c
+++ b/tools/sched_ext/scx_sdt.c
@@ -51,6 +51,7 @@ int main(int argc, char **argv)
 	signal(SIGINT, sigint_handler);
 	signal(SIGTERM, sigint_handler);
 restart:
+	optind = 1;
 	skel = SCX_OPS_OPEN(sdt_ops, scx_sdt);
 
 	while ((opt = getopt(argc, argv, "fvh")) != -1) {
diff --git a/tools/sched_ext/scx_simple.c b/tools/sched_ext/scx_simple.c
index 06d4b13bf76b..c3b48611712b 100644
--- a/tools/sched_ext/scx_simple.c
+++ b/tools/sched_ext/scx_simple.c
@@ -71,6 +71,7 @@ int main(int argc, char **argv)
 	signal(SIGINT, sigint_handler);
 	signal(SIGTERM, sigint_handler);
 restart:
+	optind = 1;
 	skel = SCX_OPS_OPEN(simple_ops, scx_simple);
 
 	while ((opt = getopt(argc, argv, "fvh")) != -1) {
diff --git a/tools/sched_ext/scx_userland.c b/tools/sched_ext/scx_userland.c
index 10b31020f44f..3f2aba658b4a 100644
--- a/tools/sched_ext/scx_userland.c
+++ b/tools/sched_ext/scx_userland.c
@@ -54,6 +54,7 @@ static bool verbose;
 static volatile int exit_req;
 static int enqueued_fd, dispatched_fd;
 
+static pthread_t stats_printer;
 static struct scx_userland *skel;
 static struct bpf_link *ops_link;
 
@@ -156,9 +157,9 @@ static int dispatch_task(__s32 pid)
 
 	err = bpf_map_update_elem(dispatched_fd, NULL, &pid, 0);
 	if (err) {
-		nr_vruntime_failed++;
+		__atomic_add_fetch(&nr_vruntime_failed, 1, __ATOMIC_RELAXED);
 	} else {
-		nr_vruntime_dispatches++;
+		__atomic_add_fetch(&nr_vruntime_dispatches, 1, __ATOMIC_RELAXED);
 	}
 
 	return err;
@@ -201,8 +202,8 @@ static int vruntime_enqueue(const struct scx_userland_enqueued_task *bpf_task)
 		return ENOENT;
 
 	update_enqueued(curr, bpf_task);
-	nr_vruntime_enqueues++;
-	nr_curr_enqueued++;
+	__atomic_add_fetch(&nr_vruntime_enqueues, 1, __ATOMIC_RELAXED);
+	__atomic_add_fetch(&nr_curr_enqueued, 1, __ATOMIC_RELAXED);
 
 	/*
 	 * Enqueue the task in a vruntime-sorted list. A more optimal data
@@ -278,9 +279,9 @@ static void dispatch_batch(void)
 			LIST_INSERT_HEAD(&vruntime_head, task, entries);
 			break;
 		}
-		nr_curr_enqueued--;
+		__atomic_sub_fetch(&nr_curr_enqueued, 1, __ATOMIC_RELAXED);
 	}
-	skel->bss->nr_scheduled = nr_curr_enqueued;
+	skel->bss->nr_scheduled = __atomic_load_n(&nr_curr_enqueued, __ATOMIC_RELAXED);
 }
 
 static void *run_stats_printer(void *arg)
@@ -305,9 +306,9 @@ static void *run_stats_printer(void *arg)
 		printf("|-----------------------|\n");
 		printf("| VRUNTIME / USER       |\n");
 		printf("|-----------------------|\n");
-		printf("|  enq:      %10llu |\n", nr_vruntime_enqueues);
-		printf("|  disp:     %10llu |\n", nr_vruntime_dispatches);
-		printf("|  failed:   %10llu |\n", nr_vruntime_failed);
+		printf("|  enq:      %10llu |\n", __atomic_load_n(&nr_vruntime_enqueues, __ATOMIC_RELAXED));
+		printf("|  disp:     %10llu |\n", __atomic_load_n(&nr_vruntime_dispatches, __ATOMIC_RELAXED));
+		printf("|  failed:   %10llu |\n", __atomic_load_n(&nr_vruntime_failed, __ATOMIC_RELAXED));
 		printf("o-----------------------o\n");
 		printf("\n\n");
 		fflush(stdout);
@@ -319,8 +320,6 @@ static void *run_stats_printer(void *arg)
 
 static int spawn_stats_thread(void)
 {
-	pthread_t stats_printer;
-
 	return pthread_create(&stats_printer, NULL, run_stats_printer, NULL);
 }
 
@@ -375,6 +374,15 @@ static void pre_bootstrap(int argc, char **argv)
 
 static void bootstrap(char *comm)
 {
+	exit_req = 0;
+	min_vruntime = 0.0;
+	__atomic_store_n(&nr_vruntime_enqueues, 0, __ATOMIC_RELAXED);
+	__atomic_store_n(&nr_vruntime_dispatches, 0, __ATOMIC_RELAXED);
+	__atomic_store_n(&nr_vruntime_failed, 0, __ATOMIC_RELAXED);
+	__atomic_store_n(&nr_curr_enqueued, 0, __ATOMIC_RELAXED);
+	memset(tasks, 0, pid_max * sizeof(*tasks));
+	LIST_INIT(&vruntime_head);
+
 	skel = SCX_OPS_OPEN(userland_ops, scx_userland);
 
 	skel->rodata->num_possible_cpus = libbpf_num_possible_cpus();
@@ -428,6 +436,7 @@ restart:
 
 	exit_req = 1;
 	bpf_link__destroy(ops_link);
+	pthread_join(stats_printer, NULL);
 	ecode = UEI_REPORT(skel, uei);
 	scx_userland__destroy(skel);
 
diff --git a/tools/testing/selftests/sched_ext/rt_stall.c b/tools/testing/selftests/sched_ext/rt_stall.c
index 015200f80f6e..ab772e336f86 100644
--- a/tools/testing/selftests/sched_ext/rt_stall.c
+++ b/tools/testing/selftests/sched_ext/rt_stall.c
@@ -23,6 +23,30 @@
 #define CORE_ID		0	/* CPU to pin tasks to */
 #define RUN_TIME        5	/* How long to run the test in seconds */
 
+/* Signal the parent that setup is complete by writing to a pipe */
+static void signal_ready(int fd)
+{
+	char c = 1;
+
+	if (write(fd, &c, 1) != 1) {
+		perror("write to ready pipe");
+		exit(EXIT_FAILURE);
+	}
+	close(fd);
+}
+
+/* Wait for a child to signal readiness via a pipe */
+static void wait_ready(int fd)
+{
+	char c;
+
+	if (read(fd, &c, 1) != 1) {
+		perror("read from ready pipe");
+		exit(EXIT_FAILURE);
+	}
+	close(fd);
+}
+
 /* Simple busy-wait function for test tasks */
 static void process_func(void)
 {
@@ -122,14 +146,24 @@ static bool sched_stress_test(bool is_ext)
 
 	float ext_runtime, rt_runtime, actual_ratio;
 	int ext_pid, rt_pid;
+	int ext_ready[2], rt_ready[2];
 
 	ksft_print_header();
 	ksft_set_plan(1);
 
+	if (pipe(ext_ready) || pipe(rt_ready)) {
+		perror("pipe");
+		ksft_exit_fail();
+	}
+
 	/* Create and set up a EXT task */
 	ext_pid = fork();
 	if (ext_pid == 0) {
+		close(ext_ready[0]);
+		close(rt_ready[0]);
+		close(rt_ready[1]);
 		set_affinity(CORE_ID);
+		signal_ready(ext_ready[1]);
 		process_func();
 		exit(0);
 	} else if (ext_pid < 0) {
@@ -140,8 +174,12 @@ static bool sched_stress_test(bool is_ext)
 	/* Create an RT task */
 	rt_pid = fork();
 	if (rt_pid == 0) {
+		close(ext_ready[0]);
+		close(ext_ready[1]);
+		close(rt_ready[0]);
 		set_affinity(CORE_ID);
 		set_sched(SCHED_FIFO, 50);
+		signal_ready(rt_ready[1]);
 		process_func();
 		exit(0);
 	} else if (rt_pid < 0) {
@@ -149,6 +187,17 @@ static bool sched_stress_test(bool is_ext)
 		ksft_exit_fail();
 	}
 
+	/*
+	 * Wait for both children to complete their setup (affinity and
+	 * scheduling policy) before starting the measurement window.
+	 * This prevents flaky failures caused by the RT child's setup
+	 * time eating into the measurement period.
+	 */
+	close(ext_ready[1]);
+	close(rt_ready[1]);
+	wait_ready(ext_ready[0]);
+	wait_ready(rt_ready[0]);
+
 	/* Let the processes run for the specified time */
 	sleep(RUN_TIME);