Real-time operating systems (RTOS) allows us to develop complex embedded systems. By using self-contained tasks (threads) each with their own context we can implement programs with multitasking behavior using a single CPU. Passing information between these tasks (inter-task communication) is an important aspect when designing an embedded application using an RTOS.

We can say that the inter-task communication implemented in real-time embedded systems is based on one or a combination of the following mechanisms:

  • memory sharing
  • signaling mechanism
  • message passing

Memory Sharing

The first thing that comes to mind as a mechanism for passing information between different tasks is using a shared memory location. It is important that the shared memory is protected either by a mutex or semaphore (see RTOS Mutex and Semaphore Basics). A very basic example of using shared memory location is a global variable. Although there is nothing stopping us from using such a variable, it is not recommended as there are far more sophisticated ways available as a means of communication between tasks.

Signaling Mechanisms

Signaling mechanisms are a basic form of communication. They indicate the occurrence of an event and can be used for synchronization purposes. We have already covered one such mechanism called a semaphore in our previous article. Now we will present two additional signaling mechanisms that have wide use in real-time operating systems – event flags and task flags. As there are no strict naming conventions, you may find these mechanisms under slightly different names in different RTOS distributions, but the function they serve is pretty much the same.

Event Flags

Event flags are bits used to encode specific information. They are used for synchronizing tasks and communication. The grouping of individual event flags is called an event group or a signal.

Events flags can be used by tasks and by interrupt service routines (ISR). A single event flag (or a group) can be accessed by many different tasks. The most common operations that can be performed on event flags are:

  • Create/delete event flags
  • Set/clear event flags
  • Read a flag’s value
  • Wait on a flag to take a specific value

An example API for using event flags can be found in the CMSIS standard.

Fig. Event flags

Task Flags

Task flags are a special form of event flags. While event flags can be accessed by all tasks, the task flags are used for notifications to a single receiving task. The most common operations that can be performed on task flags are:

  • Set/clear flags of a specific task
  • Wait on a flag to take a specific value

An example API for using task flags can be found in the CMSIS standard, the term used there is thread flags. In FreeRTOS these types of flags are defined as task notifications.

Fig. 2 Task flags

Message Passing

We can generally define two types of message passing:

  • Direct message passing – The sender and the receiver of the messages are explicitly defined. As an example, the popular FreeRTOS has Stream & Message Buffers as primitives for direct message passing between a single writer task and a single reader task.
  • Indirect message passing – The messages are placed in structures such as message queues or mailboxes and multiple tasks have read/write access.

Message Queues

Message queues are data buffers with a finite amount of entries. Each entry can contain data of a certain size (e.g 32bits). Message queues can be used for passing data between tasks and between interrupt service routines and tasks. They are implemented as thread-safe FIFO buffers. Specific actions are defined in the RTOS in case a task tries to write to a full message queue or tries to read an empty one.

The most common operations that can be performed on message queues are:

  • Create/delete a message queue
  • Get the number of messages currently stored in the queue
  • Put a message into the queue
  • Get a message from the queue
Fig. 3 Message queue

Example: Using a message queue for buffering data

Let’s have two tasks that are operating at different speeds. We can use a message queue as a buffer if one of the tasks produces a burst of data samples and the other task has to process each data sample individually at a fixed rate.

Mailboxes

A mailbox can store a single data of specific size (e.g 32-bit variable) and can be implemented as a single-entry queue. A single mailbox can be accessed by many tasks. In some RTOS distributions, mailboxes can have more than one entry, which makes them very similar to what we described as a message queue in the previous chapter.

Typical operation that can be performed on mailbox are:

  • Create/delete a mailbox
  • Write to a mailbox
  • Read a mailbox

Was this article helpful?