selftests/bpf: Add more test cases for bpf memory allocator

Add the following 3 test cases for bpf memory allocator:
1) Do allocation in bpf program and free through map free
2) Do batch per-cpu allocation and per-cpu free in bpf program
3) Do per-cpu allocation in bpf program and free through map free

For per-cpu allocation, because per-cpu allocation can not refill timely
sometimes, so test 2) and test 3) consider it is OK for
bpf_percpu_obj_new_impl() to return NULL.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20231020133202.4043247-8-houtao@huaweicloud.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

tools/testing/selftests/bpf/prog_tests/test_bpf_ma.c

@@ -9,9 +9,10 @@
 
 #include "test_bpf_ma.skel.h"
 
-void test_test_bpf_ma(void)
+static void do_bpf_ma_test(const char *name)
 {
 	struct test_bpf_ma *skel;
+	struct bpf_program *prog;
 	struct btf *btf;
 	int i, err;
 
@@ -34,6 +35,11 @@ void test_test_bpf_ma(void)
 		skel->rodata->data_btf_ids[i] = id;
 	}
 
+	prog = bpf_object__find_program_by_name(skel->obj, name);
+	if (!ASSERT_OK_PTR(prog, "invalid prog name"))
+		goto out;
+	bpf_program__set_autoload(prog, true);
+
 	err = test_bpf_ma__load(skel);
 	if (!ASSERT_OK(err, "load"))
 		goto out;
@@ -48,3 +54,15 @@ void test_test_bpf_ma(void)
 out:
 	test_bpf_ma__destroy(skel);
 }
+
+void test_test_bpf_ma(void)
+{
+	if (test__start_subtest("batch_alloc_free"))
+		do_bpf_ma_test("test_batch_alloc_free");
+	if (test__start_subtest("free_through_map_free"))
+		do_bpf_ma_test("test_free_through_map_free");
+	if (test__start_subtest("batch_percpu_alloc_free"))
+		do_bpf_ma_test("test_batch_percpu_alloc_free");
+	if (test__start_subtest("percpu_free_through_map_free"))
+		do_bpf_ma_test("test_percpu_free_through_map_free");
+}

tools/testing/selftests/bpf/progs/test_bpf_ma.c

@@ -37,10 +37,20 @@ int pid = 0;
 		__type(key, int); \
 		__type(value, struct map_value_##_size); \
 		__uint(max_entries, 128); \
-	} array_##_size SEC(".maps");
+	} array_##_size SEC(".maps")
 
-static __always_inline void batch_alloc_free(struct bpf_map *map, unsigned int batch,
-					     unsigned int idx)
+#define DEFINE_ARRAY_WITH_PERCPU_KPTR(_size) \
+	struct map_value_percpu_##_size { \
+		struct bin_data_##_size __percpu_kptr * data; \
+	}; \
+	struct { \
+		__uint(type, BPF_MAP_TYPE_ARRAY); \
+		__type(key, int); \
+		__type(value, struct map_value_percpu_##_size); \
+		__uint(max_entries, 128); \
+	} array_percpu_##_size SEC(".maps")
+
+static __always_inline void batch_alloc(struct bpf_map *map, unsigned int batch, unsigned int idx)
 {
 	struct generic_map_value *value;
 	unsigned int i, key;
@@ -65,6 +75,14 @@ static __always_inline void batch_alloc_free(struct bpf_map *map, unsigned int b
 			return;
 		}
 	}
+}
+
+static __always_inline void batch_free(struct bpf_map *map, unsigned int batch, unsigned int idx)
+{
+	struct generic_map_value *value;
+	unsigned int i, key;
+	void *old;
+
 	for (i = 0; i < batch; i++) {
 		key = i;
 		value = bpf_map_lookup_elem(map, &key);
@@ -81,8 +99,72 @@ static __always_inline void batch_alloc_free(struct bpf_map *map, unsigned int b
 	}
 }
 
+static __always_inline void batch_percpu_alloc(struct bpf_map *map, unsigned int batch,
+					       unsigned int idx)
+{
+	struct generic_map_value *value;
+	unsigned int i, key;
+	void *old, *new;
+
+	for (i = 0; i < batch; i++) {
+		key = i;
+		value = bpf_map_lookup_elem(map, &key);
+		if (!value) {
+			err = 1;
+			return;
+		}
+		/* per-cpu allocator may not be able to refill in time */
+		new = bpf_percpu_obj_new_impl(data_btf_ids[idx], NULL);
+		if (!new)
+			continue;
+
+		old = bpf_kptr_xchg(&value->data, new);
+		if (old) {
+			bpf_percpu_obj_drop(old);
+			err = 2;
+			return;
+		}
+	}
+}
+
+static __always_inline void batch_percpu_free(struct bpf_map *map, unsigned int batch,
+					      unsigned int idx)
+{
+	struct generic_map_value *value;
+	unsigned int i, key;
+	void *old;
+
+	for (i = 0; i < batch; i++) {
+		key = i;
+		value = bpf_map_lookup_elem(map, &key);
+		if (!value) {
+			err = 3;
+			return;
+		}
+		old = bpf_kptr_xchg(&value->data, NULL);
+		if (!old)
+			continue;
+		bpf_percpu_obj_drop(old);
+	}
+}
+
+#define CALL_BATCH_ALLOC(size, batch, idx) \
+	batch_alloc((struct bpf_map *)(&array_##size), batch, idx)
+
 #define CALL_BATCH_ALLOC_FREE(size, batch, idx) \
-	batch_alloc_free((struct bpf_map *)(&array_##size), batch, idx)
+	do { \
+		batch_alloc((struct bpf_map *)(&array_##size), batch, idx); \
+		batch_free((struct bpf_map *)(&array_##size), batch, idx); \
+	} while (0)
+
+#define CALL_BATCH_PERCPU_ALLOC(size, batch, idx) \
+	batch_percpu_alloc((struct bpf_map *)(&array_percpu_##size), batch, idx)
+
+#define CALL_BATCH_PERCPU_ALLOC_FREE(size, batch, idx) \
+	do { \
+		batch_percpu_alloc((struct bpf_map *)(&array_percpu_##size), batch, idx); \
+		batch_percpu_free((struct bpf_map *)(&array_percpu_##size), batch, idx); \
+	} while (0)
 
 DEFINE_ARRAY_WITH_KPTR(8);
 DEFINE_ARRAY_WITH_KPTR(16);
@@ -97,8 +179,21 @@ DEFINE_ARRAY_WITH_KPTR(1024);
 DEFINE_ARRAY_WITH_KPTR(2048);
 DEFINE_ARRAY_WITH_KPTR(4096);
 
-SEC("fentry/" SYS_PREFIX "sys_nanosleep")
-int test_bpf_mem_alloc_free(void *ctx)
+/* per-cpu kptr doesn't support bin_data_8 which is a zero-sized array */
+DEFINE_ARRAY_WITH_PERCPU_KPTR(16);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(32);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(64);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(96);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(128);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(192);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(256);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(512);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(1024);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(2048);
+DEFINE_ARRAY_WITH_PERCPU_KPTR(4096);
+
+SEC("?fentry/" SYS_PREFIX "sys_nanosleep")
+int test_batch_alloc_free(void *ctx)
 {
 	if ((u32)bpf_get_current_pid_tgid() != pid)
 		return 0;
@@ -121,3 +216,76 @@ int test_bpf_mem_alloc_free(void *ctx)
 
 	return 0;
 }
+
+SEC("?fentry/" SYS_PREFIX "sys_nanosleep")
+int test_free_through_map_free(void *ctx)
+{
+	if ((u32)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	/* Alloc 128 8-bytes objects in batch to trigger refilling,
+	 * then free these objects through map free.
+	 */
+	CALL_BATCH_ALLOC(8, 128, 0);
+	CALL_BATCH_ALLOC(16, 128, 1);
+	CALL_BATCH_ALLOC(32, 128, 2);
+	CALL_BATCH_ALLOC(64, 128, 3);
+	CALL_BATCH_ALLOC(96, 128, 4);
+	CALL_BATCH_ALLOC(128, 128, 5);
+	CALL_BATCH_ALLOC(192, 128, 6);
+	CALL_BATCH_ALLOC(256, 128, 7);
+	CALL_BATCH_ALLOC(512, 64, 8);
+	CALL_BATCH_ALLOC(1024, 32, 9);
+	CALL_BATCH_ALLOC(2048, 16, 10);
+	CALL_BATCH_ALLOC(4096, 8, 11);
+
+	return 0;
+}
+
+SEC("?fentry/" SYS_PREFIX "sys_nanosleep")
+int test_batch_percpu_alloc_free(void *ctx)
+{
+	if ((u32)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	/* Alloc 128 16-bytes per-cpu objects in batch to trigger refilling,
+	 * then free 128 16-bytes per-cpu objects in batch to trigger freeing.
+	 */
+	CALL_BATCH_PERCPU_ALLOC_FREE(16, 128, 1);
+	CALL_BATCH_PERCPU_ALLOC_FREE(32, 128, 2);
+	CALL_BATCH_PERCPU_ALLOC_FREE(64, 128, 3);
+	CALL_BATCH_PERCPU_ALLOC_FREE(96, 128, 4);
+	CALL_BATCH_PERCPU_ALLOC_FREE(128, 128, 5);
+	CALL_BATCH_PERCPU_ALLOC_FREE(192, 128, 6);
+	CALL_BATCH_PERCPU_ALLOC_FREE(256, 128, 7);
+	CALL_BATCH_PERCPU_ALLOC_FREE(512, 64, 8);
+	CALL_BATCH_PERCPU_ALLOC_FREE(1024, 32, 9);
+	CALL_BATCH_PERCPU_ALLOC_FREE(2048, 16, 10);
+	CALL_BATCH_PERCPU_ALLOC_FREE(4096, 8, 11);
+
+	return 0;
+}
+
+SEC("?fentry/" SYS_PREFIX "sys_nanosleep")
+int test_percpu_free_through_map_free(void *ctx)
+{
+	if ((u32)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	/* Alloc 128 16-bytes per-cpu objects in batch to trigger refilling,
+	 * then free these object through map free.
+	 */
+	CALL_BATCH_PERCPU_ALLOC(16, 128, 1);
+	CALL_BATCH_PERCPU_ALLOC(32, 128, 2);
+	CALL_BATCH_PERCPU_ALLOC(64, 128, 3);
+	CALL_BATCH_PERCPU_ALLOC(96, 128, 4);
+	CALL_BATCH_PERCPU_ALLOC(128, 128, 5);
+	CALL_BATCH_PERCPU_ALLOC(192, 128, 6);
+	CALL_BATCH_PERCPU_ALLOC(256, 128, 7);
+	CALL_BATCH_PERCPU_ALLOC(512, 64, 8);
+	CALL_BATCH_PERCPU_ALLOC(1024, 32, 9);
+	CALL_BATCH_PERCPU_ALLOC(2048, 16, 10);
+	CALL_BATCH_PERCPU_ALLOC(4096, 8, 11);
+
+	return 0;
+}