The Time management functions are called from idle task whenenver there is Timer interrupt. When there is timer interrupt, the priority of the Idle task (System task) is raised to the highest (0) and scheduling returns to the counting loop of the idle task. It skips the counting loop and disables the interupt to create the critical section and increments the timer count. And, it processes the timer tasks further.
Alarm handler and Cyclic handler are just executing handler at interrupt enabled context. Cyclic handler is cyclic. Alarm is just once. But, other timeout handler is not handler. Instead, waking up tasks sleeping at various resources and system calls with various reasons.
But, why two timeout redundant values? 1) Timeout queue 2) TCB timeout
System Dispatch Service
{
If (there is system error)
return system error;
Enable dispatch;
Initializes internal strucutres;
/* Enter endless loop */
for (;;) {
This task is assigned lowest priority in the system.
Maximum priority + 1;
If (Dispatch is disabled)
Special Dispatch task = Previous running task
else
No special request
Schedule and dispatch;
if (overrun handler conndition )
Call Overrun handler
/* counter loop */
for (;;) {
count down and stay here if reached zero;
Poll for Timer interrup flag;
if (timer interrrupt flag is set )
break;
}
Get idle count;
Enter critical section;/* diable interrupt */
for (till number of pending timer interrupts reaches zero)
Update system clock;
check timer queue - Timeout check processing;
check cyclic handler management
check alarm handler management
}
}
}
Timeout check processing
void timeout_check_processing(void)
{
Take queue corresponding to lower 32 bit of system time
for (till there is valid entry in the head of the queue) {
Take taskid in the head of the queue
if (task id is zero)
return;
Take TCB
for (till an entry matching the timeout value is reached) {
if (matching) { /* ?? I think this will match always */
Remove the task from the timer queue;
Take backup of the task status
Assign the task status as READY
Assign the return code based on previous status
if (in Timer queue due to dly_tsk(DLY)) {
Assign return code as E_OK
} else {
Assign return code as E_TIMEOUT
if (waiting for any resource) {
delete from the resource's waiting queue
if (resource is Mutex) {
Get the Mutex ID
if (Mutex is INHERITENCE and
priority is higher or equal
ceiling priority)
Adjust Mutex priority
} else if (resource is Message buffer)
adjust resource allocation
} else if (resource is variable m-pool)
adjust resource allocation
}
}
if (not suspended) {
Take priority
if (priority is higher than current Application priority)
Add to delayed Dispatch
/* As this function won't call dispatch */
Add to Ready queue of the task's priority
}
temporarily open and close critical
section to open up the door for dispatch
break; /* Again go to head for check */
/* Anything might have happened when you opened up the door */
}
if (reached end of queue)
return;
move to next task in the queue chain
}
}
}
void Cyclic_handler_processing(void)
{
Take the cyclic handler queue corresponding to the system time low 32 bit
for (till valid entry in the head of the queue) {
Take the cyclic handler ID at the queue head
if (cyclic handler ID is 0)
return;
for (till an entry which matches with time stamp) {
if (deadline matches with time stamp) {
delete from cyclic handler queue
calculate new target time as it is cyclic handler
add to the new queue of new target time
temporarily enable interrupt
if (cyclic hander is ON now) {
Execute cyclic handler
}
disable interrupt again
break; /* Restart from head again */
/* anything is possible in the cycle gap */
}
if (reached end of queue)
return;
move to next cyclic handler in the same queue
}
}
}
Alarm Handler Processing
void alarm_handler_processing(void)
{
Take alarm handler queue corresponding to the system time low 32 bit
for (till valid entry in the head of the queue) {
for (till an entry which matches with time stamp) {
if (deadline matches with time stamp) {
delete from alarm handler queue
reset the alarm handler entry fields
temporarily enable interrupt
Execute alarm handler
disable interrupt again
break; /* Restart from head again */
/* anything is possible in the cycle gap */
}
if (reached end of queue)
return;
move to next alarm handler in the same queue
}
}
}
Accumulating or queuing inside system processing is not real-time. Realtime never accumulates in production center. Always, wherever there is a change in the system state, it opens the gate and put the car on the road. It is like Toyota. Never keep the doors closed.
Whether the car runs on the road or stands depends on the user program. But, kernel will pass the message always at real time.
1 comment:
Post a Comment