tests/msgq_usage: Correct memory usage for cache-incoherent platforms (7fec8b28) · Commits · Wenxi XU / zephyr

When CONFIG_KERNEL_COHERENCE=y (e.g on the various intel_adsp platforms under SMP) it's not legal to share stack memory between CPUs, because the stack is cached, and the L1 cache is incoherent. The kernel will automatically detect the mistake when the memory contains a kernel object (spinlock, IPC object, etc...). But here the test was just passing async buffers into the msgq layer, and nothing watches that. The fix is simple: make them static. Fixes #35857 Signed-off-by:

Andy Ross <andrew.j.ross@intel.com>

tests/kernel/msgq/msgq_usage/src/main.c

+5 −5

Original line number	Diff line number	Diff line
		@@ -50,7 +50,7 @@ enum message_info {

		static void service_manager_entry(void p1, void p2, void *p3)
		{
		unsigned long data[2];
		static unsigned long data[2];

		while (1) {
		k_msgq_get(&manager_q, data, K_FOREVER);
		@@ -96,7 +96,7 @@ static void start_service_manager(void)

		static void service1_entry(void p1, void p2, void *p3)
		{
		unsigned long service_data[2];
		static unsigned long service_data[2];
		struct k_msgq *client;
		int ret;

		@@ -127,7 +127,7 @@ static void service1_entry(void p1, void p2, void *p3)

		static void service2_entry(void p1, void p2, void *p3)
		{
		unsigned long service_data[2];
		static unsigned long service_data[2];
		struct k_msgq *client;
		int ret;

		@@ -174,8 +174,8 @@ static void register_service(void)

		static void client_entry(void p1, void p2, void *p3)
		{
		unsigned long client_data[2];
		unsigned long service_data[2];
		static unsigned long client_data[2];
		static unsigned long service_data[2];
		struct k_msgq *service1q;
		struct k_msgq *service2q;
		bool query_service = false;

Admin message