sched_ext: Maintain an online cid mask in the scheduler arenagrafted

Schedulers on the default cid mapping treat [0, nr_online_cids) as the
online set and restart on hotplug. Schedulers that install their own mapping
with scx_bpf_cid_override() have no way to learn which cids are online: the
count no longer identifies members and the CPU-form cpumask is unusable from
cid programs. This is an obvious hole in the cid API.

Add scx_bpf_online_cmask(), a kernel-maintained cmask in the scheduler's
arena, allocated alongside the per-CPU scratch masks and populated after the
cid mapping is finalized and before ops.init(), for child schedulers too.
The pointer stays valid through ops.exit() with no reference to take. It is
the arena offset as a void pointer, the same form struct_ops arena arguments
arrive in. The verifier types the void return as a scalar for the program's
arena cast.

The mask follows the SCX hotplug notifications: seeded from cpu_active_mask
and updated before ops.cid_online/offline() runs, so it lags cpu_online_mask
only inside a hotplug transition. Updates walk the scheduler list under the
lock that also serializes unlinking. Reads are live, not atomic snapshots.
Root initialization excludes hotplug.

v2: Reworded the getter kerneldoc (Andrea Righi).

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>

1 files changed, 245 insertions, 0 deletions

diff --git a/tools/perf/tests/maps.c b/tools/perf/tests/maps.c
new file mode 100644
index 000000000..2c05d62f4
--- /dev/null
+++ b/tools/perf/tests/maps.c
@@ -0,0 +1,245 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <inttypes.h>
+#include <linux/compiler.h>
+#include <linux/kernel.h>
+#include "tests.h"
+#include "map.h"
+#include "maps.h"
+#include "dso.h"
+#include "debug.h"
+
+struct map_def {
+	const char *name;
+	u64 start;
+	u64 end;
+};
+
+struct check_maps_cb_args {
+	struct map_def *merged;
+	unsigned int i;
+};
+
+static int check_maps_cb(struct map *map, void *data)
+{
+	struct check_maps_cb_args *args = data;
+	struct map_def *merged = &args->merged[args->i];
+
+	if (map__start(map) != merged->start ||
+	    map__end(map) != merged->end ||
+	    strcmp(dso__name(map__dso(map)), merged->name) ||
+	    refcount_read(map__refcnt(map)) != 1) {
+		return 1;
+	}
+	args->i++;
+	return 0;
+}
+
+static int failed_cb(struct map *map, void *data __maybe_unused)
+{
+	pr_debug("\tstart: %" PRIu64 " end: %" PRIu64 " name: '%s' refcnt: %d\n",
+		map__start(map),
+		map__end(map),
+		dso__name(map__dso(map)),
+		refcount_read(map__refcnt(map)));
+
+	return 0;
+}
+
+static int check_maps(struct map_def *merged, unsigned int size, struct maps *maps)
+{
+	bool failed = false;
+
+	if (maps__nr_maps(maps) != size) {
+		pr_debug("Expected %d maps, got %d", size, maps__nr_maps(maps));
+		failed = true;
+	} else {
+		struct check_maps_cb_args args = {
+			.merged = merged,
+			.i = 0,
+		};
+		failed = maps__for_each_map(maps, check_maps_cb, &args);
+	}
+	if (failed) {
+		pr_debug("Expected:\n");
+		for (unsigned int i = 0; i < size; i++) {
+			pr_debug("\tstart: %" PRIu64 " end: %" PRIu64 " name: '%s' refcnt: 1\n",
+				merged[i].start, merged[i].end, merged[i].name);
+		}
+		pr_debug("Got:\n");
+		maps__for_each_map(maps, failed_cb, NULL);
+	}
+	return failed ? TEST_FAIL : TEST_OK;
+}
+
+static int test__maps__merge_in(struct test_suite *t __maybe_unused, int subtest __maybe_unused)
+{
+	unsigned int i;
+	struct map_def bpf_progs[] = {
+		{ "bpf_prog_1", 200, 300 },
+		{ "bpf_prog_2", 500, 600 },
+		{ "bpf_prog_3", 800, 900 },
+	};
+	struct map_def merged12[] = {
+		{ "kcore1",     100,  200 },
+		{ "bpf_prog_1", 200,  300 },
+		{ "kcore1",     300,  500 },
+		{ "bpf_prog_2", 500,  600 },
+		{ "kcore1",     600,  800 },
+		{ "bpf_prog_3", 800,  900 },
+		{ "kcore1",     900, 1000 },
+	};
+	struct map_def merged3[] = {
+		{ "kcore1",      100,  200 },
+		{ "bpf_prog_1",  200,  300 },
+		{ "kcore1",      300,  500 },
+		{ "bpf_prog_2",  500,  600 },
+		{ "kcore1",      600,  800 },
+		{ "bpf_prog_3",  800,  900 },
+		{ "kcore1",      900, 1000 },
+		{ "kcore3",     1000, 1100 },
+	};
+	struct map *map_kcore1, *map_kcore2, *map_kcore3;
+	int ret;
+	struct maps *maps = maps__new(NULL);
+
+	TEST_ASSERT_VAL("failed to create maps", maps);
+
+	for (i = 0; i < ARRAY_SIZE(bpf_progs); i++) {
+		struct map *map;
+
+		map = dso__new_map(bpf_progs[i].name);
+		TEST_ASSERT_VAL("failed to create map", map);
+
+		map__set_start(map, bpf_progs[i].start);
+		map__set_end(map, bpf_progs[i].end);
+		TEST_ASSERT_VAL("failed to insert map", maps__insert(maps, map) == 0);
+		map__put(map);
+	}
+
+	map_kcore1 = dso__new_map("kcore1");
+	TEST_ASSERT_VAL("failed to create map", map_kcore1);
+
+	map_kcore2 = dso__new_map("kcore2");
+	TEST_ASSERT_VAL("failed to create map", map_kcore2);
+
+	map_kcore3 = dso__new_map("kcore3");
+	TEST_ASSERT_VAL("failed to create map", map_kcore3);
+
+	/* kcore1 map overlaps over all bpf maps */
+	map__set_start(map_kcore1, 100);
+	map__set_end(map_kcore1, 1000);
+
+	/* kcore2 map hides behind bpf_prog_2 */
+	map__set_start(map_kcore2, 550);
+	map__set_end(map_kcore2, 570);
+
+	/* kcore3 map hides behind bpf_prog_3, kcore1 and adds new map */
+	map__set_start(map_kcore3, 880);
+	map__set_end(map_kcore3, 1100);
+
+	ret = maps__merge_in(maps, map_kcore1);
+	TEST_ASSERT_VAL("failed to merge map", !ret);
+
+	ret = check_maps(merged12, ARRAY_SIZE(merged12), maps);
+	TEST_ASSERT_VAL("merge check failed", !ret);
+
+	ret = maps__merge_in(maps, map_kcore2);
+	TEST_ASSERT_VAL("failed to merge map", !ret);
+
+	ret = check_maps(merged12, ARRAY_SIZE(merged12), maps);
+	TEST_ASSERT_VAL("merge check failed", !ret);
+
+	ret = maps__merge_in(maps, map_kcore3);
+	TEST_ASSERT_VAL("failed to merge map", !ret);
+
+	ret = check_maps(merged3, ARRAY_SIZE(merged3), maps);
+	TEST_ASSERT_VAL("merge check failed", !ret);
+
+	maps__zput(maps);
+	map__zput(map_kcore1);
+	map__zput(map_kcore2);
+	map__zput(map_kcore3);
+	return TEST_OK;
+}
+
+static int test__maps__fixup_overlap_and_insert(struct test_suite *t __maybe_unused,
+						int subtest __maybe_unused)
+{
+	struct map_def initial_maps[] = {
+		{ "target_map", 1000, 2000 },
+		{ "next_map",   3000, 4000 },
+	};
+	struct map_def insert_split = { "split_map", 1400, 1600 };
+	struct map_def expected_after_split[] = {
+		{ "target_map", 1000, 1400 },
+		{ "split_map",  1400, 1600 },
+		{ "target_map", 1600, 2000 },
+		{ "next_map",   3000, 4000 },
+	};
+
+	struct map_def insert_eclipse = { "eclipse_map", 2500, 4500 };
+	struct map_def expected_final[] = {
+		{ "target_map",  1000, 1400 },
+		{ "split_map",   1400, 1600 },
+		{ "target_map",  1600, 2000 },
+		{ "eclipse_map", 2500, 4500 },
+		/* "next_map" (3000-4000) is removed */
+	};
+
+	struct map *map_split, *map_eclipse;
+	int ret;
+	unsigned int i;
+	struct maps *maps = maps__new(NULL);
+
+	TEST_ASSERT_VAL("failed to create maps", maps);
+
+	for (i = 0; i < ARRAY_SIZE(initial_maps); i++) {
+		struct map *map = dso__new_map(initial_maps[i].name);
+
+		TEST_ASSERT_VAL("failed to create map", map);
+		map__set_start(map, initial_maps[i].start);
+		map__set_end(map, initial_maps[i].end);
+		TEST_ASSERT_VAL("failed to insert map", maps__insert(maps, map) == 0);
+		map__put(map);
+	}
+
+	// Check splitting.
+	map_split = dso__new_map(insert_split.name);
+	TEST_ASSERT_VAL("failed to create split map", map_split);
+	map__set_start(map_split, insert_split.start);
+	map__set_end(map_split, insert_split.end);
+
+	ret = maps__fixup_overlap_and_insert(maps, map_split);
+	TEST_ASSERT_VAL("failed to fixup and insert split map", !ret);
+
+	map__zput(map_split);
+	ret = check_maps(expected_after_split, ARRAY_SIZE(expected_after_split), maps);
+	TEST_ASSERT_VAL("split check failed", !ret);
+
+	// Check cover 1 map with another.
+	map_eclipse = dso__new_map(insert_eclipse.name);
+	TEST_ASSERT_VAL("failed to create eclipse map", map_eclipse);
+	map__set_start(map_eclipse, insert_eclipse.start);
+	map__set_end(map_eclipse, insert_eclipse.end);
+
+	ret = maps__fixup_overlap_and_insert(maps, map_eclipse);
+	TEST_ASSERT_VAL("failed to fixup and insert eclipse map", !ret);
+
+	map__zput(map_eclipse);
+	ret = check_maps(expected_final, ARRAY_SIZE(expected_final), maps);
+	TEST_ASSERT_VAL("eclipse check failed", !ret);
+
+	maps__zput(maps);
+	return TEST_OK;
+}
+
+static struct test_case tests__maps[] = {
+	TEST_CASE("Test merge_in interface", maps__merge_in),
+	TEST_CASE("Test fix up overlap interface", maps__fixup_overlap_and_insert),
+	{	.name = NULL, }
+};
+
+struct test_suite suite__maps = {
+	.desc = "Maps - per process mmaps abstraction",
+	.test_cases = tests__maps,
+};