/* * Copyright (c) 2006-2024 RT-Thread Development Team * * SPDX-License-Identifier: Apache-2.0 * * Change Logs: * Date Author Notes * 2024/10/28 Shell Added smp.smoke * 2025/12/3 ChuanN-sudo add standardized utest documentation block */ /** * Test Case Name: SMP Call Smoke 002 Test * * Test Objectives: * - Validate thread-safety of rt_smp_call_cpu_mask() under concurrent calls from multiple CPUs. * - Verify system stability with simultaneous blocking SMP calls. * - Test core APIs: rt_smp_call_cpu_mask(). * * Test Scenarios: * - One worker thread pinned to each CPU core. * - All threads simultaneously call rt_smp_call_cpu_mask() with random targets. * - Atomic counter tracks callbacks executed per thread. * * Verification Metrics: * - Callback count must match requested target count. * - Callbacks execute in interrupt-disabled context. * - All threads successfully created and pinned. * - Each thread tests all available CPUs. * * Dependencies: * - Hardware requirements: QEMU emulator or any multi-core hardware platform that supports RT-Thread. * - Software configuration: * - RT_USING_UTEST must be enabled (select "RT-Thread Utestcases" in menuconfig). * - RT_UTEST_SMP_CALL_FUNC must be enabled(enable via: RT-Thread Utestcases -> Kernel Components -> Drivers -> SMP-Call Test -> SMP-Call Smoke Test). * - Environmental Assumptions: System scheduler and SMP services working normally. * * Expected Results: * - Progress logs: A series of '#' characters indicating callback executions. * - Final output: "[ PASSED ] [ result ] testcase (components.drivers.smp_call.smoke_002)" */ #include #include #include #include #define PERCPU_TEST_COUNT 10000 #define NEWLINE_ON 80 static struct rt_semaphore _utestd_exited; static rt_thread_t _utestd[RT_CPUS_NR]; static rt_atomic_t _entry_counts[RT_CPUS_NR]; static void _logging_progress(void) { static rt_atomic_t counts; rt_ubase_t old; rt_kputs("#"); old = rt_atomic_add(&counts, 1); if (old % NEWLINE_ON == 0) { rt_kputs("\n"); } } static void _test_smp_cb(void *param) { rt_ubase_t req_cpuid = (rt_ubase_t)param; if (!rt_hw_interrupt_is_disabled()) { /* SYNC.004 */ uassert_true(0); } _logging_progress(); rt_atomic_add(&_entry_counts[req_cpuid], 1); } static void _utestd_entry(void *oncpu_param) { rt_ubase_t oncpu = (rt_ubase_t)oncpu_param; volatile int cpu_mask; volatile int popcount = 0; rt_ubase_t tested_cpus = 0; if (rt_hw_cpu_id() != oncpu) { /* SYNC.004 */ uassert_true(0); } for (size_t i = 0; i < PERCPU_TEST_COUNT; i++) { cpu_mask = rand() % RT_ALL_CPU; tested_cpus |= cpu_mask; rt_smp_call_cpu_mask(cpu_mask, _test_smp_cb, oncpu_param, SMP_CALL_WAIT_ALL); popcount += __builtin_popcount(cpu_mask); } LOG_D("popcount %d, _entry_counts[%d] %d", popcount, oncpu, _entry_counts[oncpu]); /* TARG.001 */ uassert_true(popcount == rt_atomic_load(&_entry_counts[oncpu])); /* TOP.001, TOP.002 */ uassert_true(tested_cpus == RT_ALL_CPU); rt_sem_release(&_utestd_exited); } static void _blocking_mtsafe_call(void) { rt_err_t error; for (size_t i = 0; i < RT_CPUS_NR; i++) { error = rt_thread_startup(_utestd[i]); /* SYNC.001, SYNC.002, SYNC.003 */ uassert_true(!error); } for (size_t i = 0; i < RT_CPUS_NR; i++) { rt_sem_take(&_utestd_exited, RT_WAITING_FOREVER); } } static rt_err_t utest_tc_init(void) { for (size_t i = 0; i < RT_CPUS_NR; i++) { rt_atomic_store(&_entry_counts[i], 0); _utestd[i] = rt_thread_create("utestd", _utestd_entry, (void *)i, UTEST_THR_STACK_SIZE, UTEST_THR_PRIORITY, 20); rt_thread_control(_utestd[i], RT_THREAD_CTRL_BIND_CPU, (void *)i); /* SYNC.001, SYNC.002, SYNC.003 */ uassert_true(_utestd[i] != RT_NULL); } rt_sem_init(&_utestd_exited, "utestd", 0, RT_IPC_FLAG_PRIO); srand(rt_tick_get()); return RT_EOK; } static rt_err_t utest_tc_cleanup(void) { rt_sem_detach(&_utestd_exited); return RT_EOK; } static void _testcase(void) { UTEST_UNIT_RUN(_blocking_mtsafe_call); } UTEST_TC_EXPORT(_testcase, "components.drivers.smp_call.smoke_002", utest_tc_init, utest_tc_cleanup, 10);