selftests/bpf: Add selftest suite for cpumask kfuncs
authorDavid Vernet <void@manifault.com>
Wed, 25 Jan 2023 14:38:13 +0000 (08:38 -0600)
committerAlexei Starovoitov <ast@kernel.org>
Wed, 25 Jan 2023 15:57:50 +0000 (07:57 -0800)
A recent patch added a new set of kfuncs for allocating, freeing,
manipulating, and querying cpumasks. This patch adds a new 'cpumask'
selftest suite which verifies their behavior.

Signed-off-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/r/20230125143816.721952-5-void@manifault.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
tools/testing/selftests/bpf/DENYLIST.s390x
tools/testing/selftests/bpf/prog_tests/cpumask.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/cpumask_common.h [new file with mode: 0644]
tools/testing/selftests/bpf/progs/cpumask_failure.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/cpumask_success.c [new file with mode: 0644]

index f18e367ca4b96d6d5cbf946ed9798e6b8ea835af..50924611e5bb828a633dcf30cdb21f5ebcd2f34b 100644 (file)
@@ -13,6 +13,7 @@ cgroup_hierarchical_stats                # JIT does not support calling kernel f
 cgrp_kfunc                               # JIT does not support calling kernel function
 cgrp_local_storage                       # prog_attach unexpected error: -524                                          (trampoline)
 core_read_macros                         # unknown func bpf_probe_read#4                                               (overlapping)
+cpumask                                  # JIT does not support calling kernel function
 d_path                                   # failed to auto-attach program 'prog_stat': -524                             (trampoline)
 decap_sanity                             # JIT does not support calling kernel function                                (kfunc)
 deny_namespace                           # failed to attach: ERROR: strerror_r(-524)=22                                (trampoline)
diff --git a/tools/testing/selftests/bpf/prog_tests/cpumask.c b/tools/testing/selftests/bpf/prog_tests/cpumask.c
new file mode 100644 (file)
index 0000000..5fbe457
--- /dev/null
@@ -0,0 +1,74 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2023 Meta Platforms, Inc. and affiliates. */
+
+#include <test_progs.h>
+#include "cpumask_failure.skel.h"
+#include "cpumask_success.skel.h"
+
+static const char * const cpumask_success_testcases[] = {
+       "test_alloc_free_cpumask",
+       "test_set_clear_cpu",
+       "test_setall_clear_cpu",
+       "test_first_firstzero_cpu",
+       "test_test_and_set_clear",
+       "test_and_or_xor",
+       "test_intersects_subset",
+       "test_copy_any_anyand",
+       "test_insert_leave",
+       "test_insert_remove_release",
+       "test_insert_kptr_get_release",
+};
+
+static void verify_success(const char *prog_name)
+{
+       struct cpumask_success *skel;
+       struct bpf_program *prog;
+       struct bpf_link *link = NULL;
+       pid_t child_pid;
+       int status;
+
+       skel = cpumask_success__open();
+       if (!ASSERT_OK_PTR(skel, "cpumask_success__open"))
+               return;
+
+       skel->bss->pid = getpid();
+       skel->bss->nr_cpus = libbpf_num_possible_cpus();
+
+       cpumask_success__load(skel);
+       if (!ASSERT_OK_PTR(skel, "cpumask_success__load"))
+               goto cleanup;
+
+       prog = bpf_object__find_program_by_name(skel->obj, prog_name);
+       if (!ASSERT_OK_PTR(prog, "bpf_object__find_program_by_name"))
+               goto cleanup;
+
+       link = bpf_program__attach(prog);
+       if (!ASSERT_OK_PTR(link, "bpf_program__attach"))
+               goto cleanup;
+
+       child_pid = fork();
+       if (!ASSERT_GT(child_pid, -1, "child_pid"))
+               goto cleanup;
+       if (child_pid == 0)
+               _exit(0);
+       waitpid(child_pid, &status, 0);
+       ASSERT_OK(skel->bss->err, "post_wait_err");
+
+cleanup:
+       bpf_link__destroy(link);
+       cpumask_success__destroy(skel);
+}
+
+void test_cpumask(void)
+{
+       int i;
+
+       for (i = 0; i < ARRAY_SIZE(cpumask_success_testcases); i++) {
+               if (!test__start_subtest(cpumask_success_testcases[i]))
+                       continue;
+
+               verify_success(cpumask_success_testcases[i]);
+       }
+
+       RUN_TESTS(cpumask_failure);
+}
diff --git a/tools/testing/selftests/bpf/progs/cpumask_common.h b/tools/testing/selftests/bpf/progs/cpumask_common.h
new file mode 100644 (file)
index 0000000..ad34f3b
--- /dev/null
@@ -0,0 +1,114 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2023 Meta Platforms, Inc. and affiliates. */
+
+#ifndef _CPUMASK_COMMON_H
+#define _CPUMASK_COMMON_H
+
+#include "errno.h"
+#include <stdbool.h>
+
+int err;
+
+struct __cpumask_map_value {
+       struct bpf_cpumask __kptr_ref * cpumask;
+};
+
+struct array_map {
+       __uint(type, BPF_MAP_TYPE_ARRAY);
+       __type(key, int);
+       __type(value, struct __cpumask_map_value);
+       __uint(max_entries, 1);
+} __cpumask_map SEC(".maps");
+
+struct bpf_cpumask *bpf_cpumask_create(void) __ksym;
+void bpf_cpumask_release(struct bpf_cpumask *cpumask) __ksym;
+struct bpf_cpumask *bpf_cpumask_acquire(struct bpf_cpumask *cpumask) __ksym;
+struct bpf_cpumask *bpf_cpumask_kptr_get(struct bpf_cpumask **cpumask) __ksym;
+u32 bpf_cpumask_first(const struct cpumask *cpumask) __ksym;
+u32 bpf_cpumask_first_zero(const struct cpumask *cpumask) __ksym;
+void bpf_cpumask_set_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_clear_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+bool bpf_cpumask_test_cpu(u32 cpu, const struct cpumask *cpumask) __ksym;
+bool bpf_cpumask_test_and_set_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+bool bpf_cpumask_test_and_clear_cpu(u32 cpu, struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_setall(struct bpf_cpumask *cpumask) __ksym;
+void bpf_cpumask_clear(struct bpf_cpumask *cpumask) __ksym;
+bool bpf_cpumask_and(struct bpf_cpumask *cpumask,
+                    const struct cpumask *src1,
+                    const struct cpumask *src2) __ksym;
+void bpf_cpumask_or(struct bpf_cpumask *cpumask,
+                   const struct cpumask *src1,
+                   const struct cpumask *src2) __ksym;
+void bpf_cpumask_xor(struct bpf_cpumask *cpumask,
+                    const struct cpumask *src1,
+                    const struct cpumask *src2) __ksym;
+bool bpf_cpumask_equal(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+bool bpf_cpumask_intersects(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+bool bpf_cpumask_subset(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+bool bpf_cpumask_empty(const struct cpumask *cpumask) __ksym;
+bool bpf_cpumask_full(const struct cpumask *cpumask) __ksym;
+void bpf_cpumask_copy(struct bpf_cpumask *dst, const struct cpumask *src) __ksym;
+u32 bpf_cpumask_any(const struct cpumask *src) __ksym;
+u32 bpf_cpumask_any_and(const struct cpumask *src1, const struct cpumask *src2) __ksym;
+
+static inline const struct cpumask *cast(struct bpf_cpumask *cpumask)
+{
+       return (const struct cpumask *)cpumask;
+}
+
+static inline struct bpf_cpumask *create_cpumask(void)
+{
+       struct bpf_cpumask *cpumask;
+
+       cpumask = bpf_cpumask_create();
+       if (!cpumask) {
+               err = 1;
+               return NULL;
+       }
+
+       if (!bpf_cpumask_empty(cast(cpumask))) {
+               err = 2;
+               bpf_cpumask_release(cpumask);
+               return NULL;
+       }
+
+       return cpumask;
+}
+
+static inline struct __cpumask_map_value *cpumask_map_value_lookup(void)
+{
+       u32 key = 0;
+
+       return bpf_map_lookup_elem(&__cpumask_map, &key);
+}
+
+static inline int cpumask_map_insert(struct bpf_cpumask *mask)
+{
+       struct __cpumask_map_value local, *v;
+       long status;
+       struct bpf_cpumask *old;
+       u32 key = 0;
+
+       local.cpumask = NULL;
+       status = bpf_map_update_elem(&__cpumask_map, &key, &local, 0);
+       if (status) {
+               bpf_cpumask_release(mask);
+               return status;
+       }
+
+       v = bpf_map_lookup_elem(&__cpumask_map, &key);
+       if (!v) {
+               bpf_cpumask_release(mask);
+               return -ENOENT;
+       }
+
+       old = bpf_kptr_xchg(&v->cpumask, mask);
+       if (old) {
+               bpf_cpumask_release(old);
+               return -EEXIST;
+       }
+
+       return 0;
+}
+
+#endif /* _CPUMASK_COMMON_H */
diff --git a/tools/testing/selftests/bpf/progs/cpumask_failure.c b/tools/testing/selftests/bpf/progs/cpumask_failure.c
new file mode 100644 (file)
index 0000000..33e8e86
--- /dev/null
@@ -0,0 +1,126 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2023 Meta Platforms, Inc. and affiliates. */
+
+#include <vmlinux.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_helpers.h>
+#include "bpf_misc.h"
+
+#include "cpumask_common.h"
+
+char _license[] SEC("license") = "GPL";
+
+/* Prototype for all of the program trace events below:
+ *
+ * TRACE_EVENT(task_newtask,
+ *         TP_PROTO(struct task_struct *p, u64 clone_flags)
+ */
+
+SEC("tp_btf/task_newtask")
+__failure __msg("Unreleased reference")
+int BPF_PROG(test_alloc_no_release, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       cpumask = create_cpumask();
+
+       /* cpumask is never released. */
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+__failure __msg("NULL pointer passed to trusted arg0")
+int BPF_PROG(test_alloc_double_release, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       cpumask = create_cpumask();
+
+       /* cpumask is released twice. */
+       bpf_cpumask_release(cpumask);
+       bpf_cpumask_release(cpumask);
+
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+__failure __msg("bpf_cpumask_acquire args#0 expected pointer to STRUCT bpf_cpumask")
+int BPF_PROG(test_acquire_wrong_cpumask, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       /* Can't acquire a non-struct bpf_cpumask. */
+       cpumask = bpf_cpumask_acquire((struct bpf_cpumask *)task->cpus_ptr);
+
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+__failure __msg("bpf_cpumask_set_cpu args#1 expected pointer to STRUCT bpf_cpumask")
+int BPF_PROG(test_mutate_cpumask, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       /* Can't set the CPU of a non-struct bpf_cpumask. */
+       bpf_cpumask_set_cpu(0, (struct bpf_cpumask *)task->cpus_ptr);
+
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+__failure __msg("Unreleased reference")
+int BPF_PROG(test_insert_remove_no_release, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+       struct __cpumask_map_value *v;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (cpumask_map_insert(cpumask))
+               return 0;
+
+       v = cpumask_map_value_lookup();
+       if (!v)
+               return 0;
+
+       cpumask = bpf_kptr_xchg(&v->cpumask, NULL);
+
+       /* cpumask is never released. */
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+__failure __msg("Unreleased reference")
+int BPF_PROG(test_kptr_get_no_release, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+       struct __cpumask_map_value *v;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (cpumask_map_insert(cpumask))
+               return 0;
+
+       v = cpumask_map_value_lookup();
+       if (!v)
+               return 0;
+
+       cpumask = bpf_cpumask_kptr_get(&v->cpumask);
+
+       /* cpumask is never released. */
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+__failure __msg("NULL pointer passed to trusted arg0")
+int BPF_PROG(test_cpumask_null, struct task_struct *task, u64 clone_flags)
+{
+  /* NULL passed to KF_TRUSTED_ARGS kfunc. */
+       bpf_cpumask_empty(NULL);
+
+       return 0;
+}
diff --git a/tools/testing/selftests/bpf/progs/cpumask_success.c b/tools/testing/selftests/bpf/progs/cpumask_success.c
new file mode 100644 (file)
index 0000000..1d38bc6
--- /dev/null
@@ -0,0 +1,426 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2023 Meta Platforms, Inc. and affiliates. */
+
+#include <vmlinux.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_helpers.h>
+
+#include "cpumask_common.h"
+
+char _license[] SEC("license") = "GPL";
+
+int pid, nr_cpus;
+
+static bool is_test_task(void)
+{
+       int cur_pid = bpf_get_current_pid_tgid() >> 32;
+
+       return pid == cur_pid;
+}
+
+static bool create_cpumask_set(struct bpf_cpumask **out1,
+                              struct bpf_cpumask **out2,
+                              struct bpf_cpumask **out3,
+                              struct bpf_cpumask **out4)
+{
+       struct bpf_cpumask *mask1, *mask2, *mask3, *mask4;
+
+       mask1 = create_cpumask();
+       if (!mask1)
+               return false;
+
+       mask2 = create_cpumask();
+       if (!mask2) {
+               bpf_cpumask_release(mask1);
+               err = 3;
+               return false;
+       }
+
+       mask3 = create_cpumask();
+       if (!mask3) {
+               bpf_cpumask_release(mask1);
+               bpf_cpumask_release(mask2);
+               err = 4;
+               return false;
+       }
+
+       mask4 = create_cpumask();
+       if (!mask4) {
+               bpf_cpumask_release(mask1);
+               bpf_cpumask_release(mask2);
+               bpf_cpumask_release(mask3);
+               err = 5;
+               return false;
+       }
+
+       *out1 = mask1;
+       *out2 = mask2;
+       *out3 = mask3;
+       *out4 = mask4;
+
+       return true;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_alloc_free_cpumask, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       if (!is_test_task())
+               return 0;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       bpf_cpumask_release(cpumask);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_set_clear_cpu, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       if (!is_test_task())
+               return 0;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       bpf_cpumask_set_cpu(0, cpumask);
+       if (!bpf_cpumask_test_cpu(0, cast(cpumask))) {
+               err = 3;
+               goto release_exit;
+       }
+
+       bpf_cpumask_clear_cpu(0, cpumask);
+       if (bpf_cpumask_test_cpu(0, cast(cpumask))) {
+               err = 4;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(cpumask);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_setall_clear_cpu, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       if (!is_test_task())
+               return 0;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       bpf_cpumask_setall(cpumask);
+       if (!bpf_cpumask_full(cast(cpumask))) {
+               err = 3;
+               goto release_exit;
+       }
+
+       bpf_cpumask_clear(cpumask);
+       if (!bpf_cpumask_empty(cast(cpumask))) {
+               err = 4;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(cpumask);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_first_firstzero_cpu, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       if (!is_test_task())
+               return 0;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (bpf_cpumask_first(cast(cpumask)) < nr_cpus) {
+               err = 3;
+               goto release_exit;
+       }
+
+       if (bpf_cpumask_first_zero(cast(cpumask)) != 0) {
+               bpf_printk("first zero: %d", bpf_cpumask_first_zero(cast(cpumask)));
+               err = 4;
+               goto release_exit;
+       }
+
+       bpf_cpumask_set_cpu(0, cpumask);
+       if (bpf_cpumask_first(cast(cpumask)) != 0) {
+               err = 5;
+               goto release_exit;
+       }
+
+       if (bpf_cpumask_first_zero(cast(cpumask)) != 1) {
+               err = 6;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(cpumask);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_test_and_set_clear, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+
+       if (!is_test_task())
+               return 0;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (bpf_cpumask_test_and_set_cpu(0, cpumask)) {
+               err = 3;
+               goto release_exit;
+       }
+
+       if (!bpf_cpumask_test_and_set_cpu(0, cpumask)) {
+               err = 4;
+               goto release_exit;
+       }
+
+       if (!bpf_cpumask_test_and_clear_cpu(0, cpumask)) {
+               err = 5;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(cpumask);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_and_or_xor, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *mask1, *mask2, *dst1, *dst2;
+
+       if (!is_test_task())
+               return 0;
+
+       if (!create_cpumask_set(&mask1, &mask2, &dst1, &dst2))
+               return 0;
+
+       bpf_cpumask_set_cpu(0, mask1);
+       bpf_cpumask_set_cpu(1, mask2);
+
+       if (bpf_cpumask_and(dst1, cast(mask1), cast(mask2))) {
+               err = 6;
+               goto release_exit;
+       }
+       if (!bpf_cpumask_empty(cast(dst1))) {
+               err = 7;
+               goto release_exit;
+       }
+
+       bpf_cpumask_or(dst1, cast(mask1), cast(mask2));
+       if (!bpf_cpumask_test_cpu(0, cast(dst1))) {
+               err = 8;
+               goto release_exit;
+       }
+       if (!bpf_cpumask_test_cpu(1, cast(dst1))) {
+               err = 9;
+               goto release_exit;
+       }
+
+       bpf_cpumask_xor(dst2, cast(mask1), cast(mask2));
+       if (!bpf_cpumask_equal(cast(dst1), cast(dst2))) {
+               err = 10;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(mask1);
+       bpf_cpumask_release(mask2);
+       bpf_cpumask_release(dst1);
+       bpf_cpumask_release(dst2);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_intersects_subset, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *mask1, *mask2, *dst1, *dst2;
+
+       if (!is_test_task())
+               return 0;
+
+       if (!create_cpumask_set(&mask1, &mask2, &dst1, &dst2))
+               return 0;
+
+       bpf_cpumask_set_cpu(0, mask1);
+       bpf_cpumask_set_cpu(1, mask2);
+       if (bpf_cpumask_intersects(cast(mask1), cast(mask2))) {
+               err = 6;
+               goto release_exit;
+       }
+
+       bpf_cpumask_or(dst1, cast(mask1), cast(mask2));
+       if (!bpf_cpumask_subset(cast(mask1), cast(dst1))) {
+               err = 7;
+               goto release_exit;
+       }
+
+       if (!bpf_cpumask_subset(cast(mask2), cast(dst1))) {
+               err = 8;
+               goto release_exit;
+       }
+
+       if (bpf_cpumask_subset(cast(dst1), cast(mask1))) {
+               err = 9;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(mask1);
+       bpf_cpumask_release(mask2);
+       bpf_cpumask_release(dst1);
+       bpf_cpumask_release(dst2);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_copy_any_anyand, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *mask1, *mask2, *dst1, *dst2;
+       u32 cpu;
+
+       if (!is_test_task())
+               return 0;
+
+       if (!create_cpumask_set(&mask1, &mask2, &dst1, &dst2))
+               return 0;
+
+       bpf_cpumask_set_cpu(0, mask1);
+       bpf_cpumask_set_cpu(1, mask2);
+       bpf_cpumask_or(dst1, cast(mask1), cast(mask2));
+
+       cpu = bpf_cpumask_any(cast(mask1));
+       if (cpu != 0) {
+               err = 6;
+               goto release_exit;
+       }
+
+       cpu = bpf_cpumask_any(cast(dst2));
+       if (cpu < nr_cpus) {
+               err = 7;
+               goto release_exit;
+       }
+
+       bpf_cpumask_copy(dst2, cast(dst1));
+       if (!bpf_cpumask_equal(cast(dst1), cast(dst2))) {
+               err = 8;
+               goto release_exit;
+       }
+
+       cpu = bpf_cpumask_any(cast(dst2));
+       if (cpu > 1) {
+               err = 9;
+               goto release_exit;
+       }
+
+       cpu = bpf_cpumask_any_and(cast(mask1), cast(mask2));
+       if (cpu < nr_cpus) {
+               err = 10;
+               goto release_exit;
+       }
+
+release_exit:
+       bpf_cpumask_release(mask1);
+       bpf_cpumask_release(mask2);
+       bpf_cpumask_release(dst1);
+       bpf_cpumask_release(dst2);
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_insert_leave, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+       struct __cpumask_map_value *v;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (cpumask_map_insert(cpumask))
+               err = 3;
+
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_insert_remove_release, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+       struct __cpumask_map_value *v;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (cpumask_map_insert(cpumask)) {
+               err = 3;
+               return 0;
+       }
+
+       v = cpumask_map_value_lookup();
+       if (!v) {
+               err = 4;
+               return 0;
+       }
+
+       cpumask = bpf_kptr_xchg(&v->cpumask, NULL);
+       if (cpumask)
+               bpf_cpumask_release(cpumask);
+       else
+               err = 5;
+
+       return 0;
+}
+
+SEC("tp_btf/task_newtask")
+int BPF_PROG(test_insert_kptr_get_release, struct task_struct *task, u64 clone_flags)
+{
+       struct bpf_cpumask *cpumask;
+       struct __cpumask_map_value *v;
+
+       cpumask = create_cpumask();
+       if (!cpumask)
+               return 0;
+
+       if (cpumask_map_insert(cpumask)) {
+               err = 3;
+               return 0;
+       }
+
+       v = cpumask_map_value_lookup();
+       if (!v) {
+               err = 4;
+               return 0;
+       }
+
+       cpumask = bpf_cpumask_kptr_get(&v->cpumask);
+       if (cpumask)
+               bpf_cpumask_release(cpumask);
+       else
+               err = 5;
+
+       return 0;
+}