Calling OS_CountSemTimedWait on RTEMS with 0 timeout waits forever

[skliper]

Describe the bug
See title. Also functional test at osal/src/tests/count-sem-timeout-test.c is missing a 0 timeout test, or this would have been caught.

From https://docs.rtems.org/doxygen/branches/master/group__ClassicSem.html:
rtems_semaphore_obtain - timeout of 0 means wait forever when RTEMS_WAIT option is passed in

To Reproduce
Observed using CF on RTEMS, the flow control semaphore hangs CF if not given. Could reproduce by adding a 0 timeout case to osal/src/tests/count-sem-timeout-test.c.

Expected behavior
Immediate return when passing 0 timeout to OS_CountSemTimedWait. Although not explicit in the OSAL API, it's implied (and how it works for POSIX):

osal/src/os/inc/osapi-countsem.h

Lines 113 to 131 in 99e3b40

	/*-------------------------------------------------------------------------------------*/
	/**
	* @brief Decrement the semaphore value with timeout
	*
	* The function locks the semaphore referenced by sem_id. However,
	* if the semaphore cannot be locked without waiting for another process
	* or thread to unlock the semaphore, this wait shall be terminated when
	* the specified timeout, msecs, expires.
	*
	* @param[in] sem_id The object ID to operate on
	* @param[in] msecs The maximum amount of time to block, in milliseconds
	*
	* @return Execution status, see @ref OSReturnCodes
	* @retval #OS_SUCCESS @copybrief OS_SUCCESS
	* @retval #OS_SEM_TIMEOUT if semaphore was not relinquished in time
	* @retval #OS_ERR_INVALID_ID if the ID passed in is not a valid semaphore ID
	* @retval #OS_SEM_FAILURE if an unspecified implementation error occurs @covtest
	*/
	int32 OS_CountSemTimedWait(osal_id_t sem_id, uint32 msecs);

Could check if TimeInTicks == 0, pass in RTEMS_NO_WAIT.

Code snips

osal/src/os/rtems/src/os-impl-countsem.c

Lines 190 to 216 in 99e3b40

	int32 OS_CountSemTimedWait_Impl(const OS_object_token_t *token, uint32 msecs)
	{
	rtems_status_code status;
	int TimeInTicks;
	OS_impl_countsem_internal_record_t *impl;
	impl = OS_OBJECT_TABLE_GET(OS_impl_count_sem_table, *token);
	if (OS_Milli2Ticks(msecs, &TimeInTicks) != OS_SUCCESS)
	{
	return OS_ERROR;
	}
	status = rtems_semaphore_obtain(impl->id, RTEMS_WAIT, TimeInTicks);
	if (status == RTEMS_TIMEOUT)
	{
	return OS_SEM_TIMEOUT;
	}
	if (status != RTEMS_SUCCESSFUL)
	{
	OS_DEBUG("Unhandled semaphore_obtain error: %s\n", rtems_status_text(status));
	return OS_SEM_FAILURE;
	}
	return OS_SUCCESS;
	}

System observed on:

• Hardware: GR712RC
• OS: RTEMS 5
• Versions: applies to main

Additional context
None.

Reporter Info
Jacob Hageman - NASA/GSFC

[skliper]

Note - also impacts OS_BinSemTimedWait

[skliper]

Note - added separate issue to cover adding functional test of 0 timeout to counting and binary semaphore logic:

• Add functional test for sem takes with 0 timeout #1411