selftests/bpf: Test the release of map btf

When there is bpf_list_head or bpf_rb_root field in map value, the free
of map btf and the free of map value may run concurrently and there may
be use-after-free problem, so add two test cases to demonstrate it. And
the use-after-free problem can been easily reproduced by using bpf_next
tree and a KASAN-enabled kernel.

The first test case tests the racing between the free of map btf and the
free of array map. It constructs the racing by releasing the array map in
the end after other ref-counter of map btf has been released. To delay
the free of array map and make it be invoked after btf_free_rcu() is
invoked, it stresses system_unbound_wq by closing multiple percpu array
maps before it closes the array map.

The second case tests the racing between the free of map btf and the
free of inner map. Beside using the similar method as the first one
does, it uses bpf_map_delete_elem() to delete the inner map and to defer
the release of inner map after one RCU grace period.

The reason for using two skeletons is to prevent the release of outer
map and inner map in map_in_map_btf.c interfering the release of bpf
map in normal_map_btf.c.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/bpf/20231216035510.4030605-1-houtao@huaweicloud.com

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

+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2023. Huawei Technologies Co., Ltd */
+#include <test_progs.h>
+
+#include "normal_map_btf.skel.h"
+#include "map_in_map_btf.skel.h"
+
+static void do_test_normal_map_btf(void)
+{
+	struct normal_map_btf *skel;
+	int i, err, new_fd = -1;
+	int map_fd_arr[64];
+
+	skel = normal_map_btf__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "open_load"))
+		return;
+
+	err = normal_map_btf__attach(skel);
+	if (!ASSERT_OK(err, "attach"))
+		goto out;
+
+	skel->bss->pid = getpid();
+	usleep(1);
+	ASSERT_TRUE(skel->bss->done, "done");
+
+	/* Use percpu_array to slow bpf_map_free_deferred() down.
+	 * The memory allocation may fail, so doesn't check the returned fd.
+	 */
+	for (i = 0; i < ARRAY_SIZE(map_fd_arr); i++)
+		map_fd_arr[i] = bpf_map_create(BPF_MAP_TYPE_PERCPU_ARRAY, NULL, 4, 4, 256, NULL);
+
+	/* Close array fd later */
+	new_fd = dup(bpf_map__fd(skel->maps.array));
+out:
+	normal_map_btf__destroy(skel);
+	if (new_fd < 0)
+		return;
+	/* Use kern_sync_rcu() to wait for the start of the free of the bpf
+	 * program and use an assumed delay to wait for the release of the map
+	 * btf which is held by other maps (e.g, bss). After that, array map
+	 * holds the last reference of map btf.
+	 */
+	kern_sync_rcu();
+	usleep(4000);
+	/* Spawn multiple kworkers to delay the invocation of
+	 * bpf_map_free_deferred() for array map.
+	 */
+	for (i = 0; i < ARRAY_SIZE(map_fd_arr); i++) {
+		if (map_fd_arr[i] < 0)
+			continue;
+		close(map_fd_arr[i]);
+	}
+	close(new_fd);
+}
+
+static void do_test_map_in_map_btf(void)
+{
+	int err, zero = 0, new_fd = -1;
+	struct map_in_map_btf *skel;
+
+	skel = map_in_map_btf__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "open_load"))
+		return;
+
+	err = map_in_map_btf__attach(skel);
+	if (!ASSERT_OK(err, "attach"))
+		goto out;
+
+	skel->bss->pid = getpid();
+	usleep(1);
+	ASSERT_TRUE(skel->bss->done, "done");
+
+	/* Close inner_array fd later */
+	new_fd = dup(bpf_map__fd(skel->maps.inner_array));
+	/* Defer the free of inner_array */
+	err = bpf_map__delete_elem(skel->maps.outer_array, &zero, sizeof(zero), 0);
+	ASSERT_OK(err, "delete inner map");
+out:
+	map_in_map_btf__destroy(skel);
+	if (new_fd < 0)
+		return;
+	/* Use kern_sync_rcu() to wait for the start of the free of the bpf
+	 * program and use an assumed delay to wait for the free of the outer
+	 * map and the release of map btf. After that, inner map holds the last
+	 * reference of map btf.
+	 */
+	kern_sync_rcu();
+	usleep(10000);
+	close(new_fd);
+}
+
+void test_map_btf(void)
+{
+	if (test__start_subtest("array_btf"))
+		do_test_normal_map_btf();
+	if (test__start_subtest("inner_array_btf"))
+		do_test_map_in_map_btf();
+}

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

+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2023. Huawei Technologies Co., Ltd */
+#include <vmlinux.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_helpers.h>
+
+#include "bpf_misc.h"
+#include "bpf_experimental.h"
+
+struct node_data {
+	__u64 data;
+	struct bpf_list_node node;
+};
+
+struct map_value {
+	struct bpf_list_head head __contains(node_data, node);
+	struct bpf_spin_lock lock;
+};
+
+struct inner_array_type {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__type(key, int);
+	__type(value, struct map_value);
+	__uint(max_entries, 1);
+} inner_array SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY_OF_MAPS);
+	__uint(key_size, 4);
+	__uint(value_size, 4);
+	__uint(max_entries, 1);
+	__array(values, struct inner_array_type);
+} outer_array SEC(".maps") = {
+	.values = {
+		[0] = &inner_array,
+	},
+};
+
+char _license[] SEC("license") = "GPL";
+
+int pid = 0;
+bool done = false;
+
+SEC("fentry/" SYS_PREFIX "sys_nanosleep")
+int add_to_list_in_inner_array(void *ctx)
+{
+	struct map_value *value;
+	struct node_data *new;
+	struct bpf_map *map;
+	int zero = 0;
+
+	if (done || (u32)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	map = bpf_map_lookup_elem(&outer_array, &zero);
+	if (!map)
+		return 0;
+
+	value = bpf_map_lookup_elem(map, &zero);
+	if (!value)
+		return 0;
+
+	new = bpf_obj_new(typeof(*new));
+	if (!new)
+		return 0;
+
+	bpf_spin_lock(&value->lock);
+	bpf_list_push_back(&value->head, &new->node);
+	bpf_spin_unlock(&value->lock);
+	done = true;
+
+	return 0;
+}

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

+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2023. Huawei Technologies Co., Ltd */
+#include <vmlinux.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_helpers.h>
+
+#include "bpf_misc.h"
+#include "bpf_experimental.h"
+
+struct node_data {
+	__u64 data;
+	struct bpf_list_node node;
+};
+
+struct map_value {
+	struct bpf_list_head head __contains(node_data, node);
+	struct bpf_spin_lock lock;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__type(key, int);
+	__type(value, struct map_value);
+	__uint(max_entries, 1);
+} array SEC(".maps");
+
+char _license[] SEC("license") = "GPL";
+
+int pid = 0;
+bool done = false;
+
+SEC("fentry/" SYS_PREFIX "sys_nanosleep")
+int add_to_list_in_array(void *ctx)
+{
+	struct map_value *value;
+	struct node_data *new;
+	int zero = 0;
+
+	if (done || (u32)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	value = bpf_map_lookup_elem(&array, &zero);
+	if (!value)
+		return 0;
+
+	new = bpf_obj_new(typeof(*new));
+	if (!new)
+		return 0;
+
+	bpf_spin_lock(&value->lock);
+	bpf_list_push_back(&value->head, &new->node);
+	bpf_spin_unlock(&value->lock);
+	done = true;
+
+	return 0;
+}