kernel/poll: no error happened when mutil-threads poll a same event at a same time.

[Zhaoningx]

Describe the bug
I took a look at Zephyr Specification https://docs.zephyrproject.org/latest/reference/kernel/other/polling.html#c.k_poll about kernel/poll , I think there is a test point that was not tested before, the info as below:

It means if there are two thread polling a same event(sem/fifo) at a same time, k_poll() immediately returns with the return value -EADDRINUSE. Is this right?
So I designed a testcase to test this, the code as below:

static K_SEM_DEFINE(test_sem, 0, 1);
static void poll_sem_t1(void *p1, void *p2, void *p3)
{
	printk("poll_sem_t1\n");

	int rc;
	struct k_poll_event *event = p1;

	rc = k_poll(event, 1, K_FOREVER);
	printk("rc1 = %d\n", rc);
	k_sem_take(&test_sem, K_FOREVER);
}

static void poll_sem_t2(void *p1, void *p2, void *p3)
{
	printk("poll_sem_t2\n");

	int rc;
	struct k_poll_event *event = p1;

	rc = k_poll(event, 1, K_FOREVER);
	printk("rc2 = %d\n", rc);
	k_sem_take(&test_sem, K_FOREVER);
}

void test_mutil_thread_poll(void)
{
	const int main_low_prio = 10;
	struct k_poll_event event;
	int old_prio = k_thread_priority_get(k_current_get());

	k_thread_priority_set(k_current_get(), main_low_prio);
	k_poll_event_init(&event, K_POLL_TYPE_SEM_AVAILABLE,
			  K_POLL_MODE_NOTIFY_ONLY, &test_sem);

	/* Spawn two threads */
	k_thread_create(&test_thread3, test_stack3,
			K_THREAD_STACK_SIZEOF(test_stack3),
			poll_sem_t1, &event, NULL, NULL, 10,
			K_INHERIT_PERMS, K_NO_WAIT);
	k_thread_create(&test_thread4, test_stack4,
			K_THREAD_STACK_SIZEOF(test_stack4),
			poll_sem_t2, &event, NULL, NULL, 10,
			K_INHERIT_PERMS, K_MSEC(10));


	k_sleep(K_MSEC(500));
	//printk("EADDRINUSE = %d\n", -EADDRINUSE);
	/* Make sem available */
	k_sem_give(&test_sem);
	k_sleep(K_MSEC(20));


	k_thread_priority_set(k_current_get(), old_prio);
}

and the output as below:

START - test_mutil_thread_poll
poll_sem_fifo_t1
poll_sem_fifo_t2
rc2 = 0
 PASS - test_mutil_thread_poll in 0.540 seconds

As you see, there is no error return value returned. So I think either the zephyr specification is wrong or there is something wrong with my code. Could you have any comments about this?

[andyross]

I have no idea what the documentation is trying to say there. This works, there is a list of pollers on each semaphore or fifo and as many events can wait in it as the app wants. It's true that they don't sort according to priority in the same way a wait queue does, so maybe that's what Ben was intending to enforce? I think we can fix this in documentation, though we should probably also add a warning that the waiters will be served in first-come-first-serve order and not prioritiy order.

[Zhaoningx]

Hi @andyross , yep, I totally agree with you, we can fix this in documentation, and would you like some help?