Enter Critical Section

From OpenMBD
< HANcoder‎ | STM32‎ | Blocks
Revision as of 09:19, 5 May 2017 by Jasonvankolfschoten (talk | contribs) (Jasonvankolfschoten moved page HANcoder/Enter Critical Section to HANcoder/STM32/Blocks/Enter Critical Section without leaving a redirect)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Supported Targets: Olimexino, E407, P405

General information about Enter Critical Section

Critical sections are needed whenever a piece of software needs to execute without being interrupted. They are typically used to guarantee data consistency. The HANcoder environment contains 2 blocks to creating critical sections: "Enter Critical Section" and "Enter Critical Section". These blocks must always be used as a pair. So for every "Enter Critical Section" block in your Simulink model, you must also add a "Exit Critical Section" block.

Critical section are only needed when signal values are shared between the model (that executes at a fixed time interval) and a function-call subsystem connected to an interrupt source (that executes at interrupt level). This could be the case when you have the "Timer Input IRQ" block in your model.

The topic of critical sections is best explained with an example. Let's assume you have 2 input signals that belong together: (1) "event_info" and (2) "event_timestamp". Furthermore, these 2 signals are inputs to a function-call subsystem that executes at interrupt level. See the following illustration:


The model in the above illustration has a data consistency problem. During software execution there is a small chance that the timer interrupt occurs after the "event_info" signal was written, but right before the write to "event_timestamp". In the function-call subsystem the "event_info" signal is up-to-date but the "event_timestamp" signal not yet. A critical section is needed to solve this data consistency problem. This is best achieved with a stateflow chart:


As a result of the critical section, the write to the 2 signals "event_info_out" and "event_timestamp_out" can not be interrupted. When these 2 signals are now used as the inputs to the function-call subsystem that executes at interrupt level, there are no more data consistency problems.