unblocking more tasks on semaphore

paha11 wrote on Sunday, October 02, 2011:

Hello,

I wonder if more tasks are waiting on a semaphore and semaphore is signaled what tasks are woken up? All or arbitary one or the first one that took the semaphore?

(Cortex-M3, freeRTOS 7.0)

thank you

rtel wrote on Sunday, October 02, 2011:

None of the above.

To make FreeRTOS real time, it has to be the highest priority task that is waiting for the semaphore, so that is what it does. 

It unblocks one task, and it is always the highest priority task that is waiting - no matter how many lower priority tasks attempted to take the semaphore first.  Any system that does not do that is basically broken as it would be designing priority inversion into the kernel - rather than trying to design it out.

Regards.

paha11 wrote on Monday, October 03, 2011:

Hi Richard

thank you for the reply.

But I could have high prio producers (all with the same prio) of some commands and low prio consumer of the commands. If all the producers fill up the freeRTOS queue they all should sleep on the queue until the consumer picks up one item. I don’t understand what happens with producers after the consumer gets one item from the queue. Could you please clear the point for me?

Thank you

richard_damon wrote on Monday, October 03, 2011:

If a producer tries to put date into a full queue, that producer will be held (unless the timeout for the queue write is 0) until there is space in the queue or the timeout expires. When a task takes an item from the queue, the highest priority task waiting to write to the queue wakes up and completes its write. The rest continue to be held.

paha11 wrote on Monday, October 03, 2011:

And if all the waiting tasks have the same priority, than the rule is the first to acquire the first to wake up?

rtel wrote on Monday, October 03, 2011:

And if all the waiting tasks have the same priority, than the rule is the first to acquire the first to wake up?

Yes

Regards.