ARINC 653 (Avionics Application Software Standard Interface) is a software specification for space and time partitioning in safety-critical avionics real-time operating systems (RTOS). It allows the hosting of multiple applications of different software levels on the same hardware in the context of an Integrated Modular Avionics architecture.
It is part of ARINC 600-Series Standards for Digital Aircraft & Flight Simulators.
Overview
In order to decouple the real-time operating system platform from the application software, ARINC 653 defines an API called APplication EXecutive (APEX).
Each application software is called a partition and has its own memory space. It also has a dedicated time slot allocated by the APEX API. Within each partition, multitasking is allowed. The APEX API provides services to manage partitions, processes and timing, as well as partition/process communication and error handling. The partitioning environment can be implemented by using a hypervisor to map partitions to virtual machines, but this is not required.
The standard is overseen by the AEEC APEX Subcommittee.
History
Initial version
The initial version of ARINC 653 was published on October 10, 1996.
ARINC 653-1
Supplement 1 was published in January 1997 and introduced the concepts of APEX and Time and Space partitioning.
ARINC 653-2
Supplement 2 was published in 3 parts between March 2006 and January 2007: Part 1 (mandatory services): ARINC 653 partition management, Cold start and warm start definition, Application software error handling, ARINC 653 compliance, Ada and C language bindings; Part 2 (optional services): File system access, Data logging, Service Access points, ... Part 3 (Conformity Test Specification);
Current Organization of Standard
Part 0 - Introduction to ARINC 653 (currently at revision 3, released November 2021)
Part 1 - Required Services (currently at revision 5, released December 2019) Part 2 - Extended Services (currently at revision 4, released December 2019) Part 3A - Conformity Test Specification for Required Services (currently at revision 2, released November 2021) Part 3B - Conformity Test Specification for Extended Services (currently at revision c1, released July 2019) Part 4 - Subset Services (currently at revision 0, released June 2012) Part 5 - Core Software Recommended Capabilities (currently at revision 1, released August 2019)
Basic principles of partitioning
ARINC 653 Platform
An ARINC 653 platform contains: A hardware platform allowing real-time computing deterministic services. An abstraction layer managing the timer and space partitioning constraints of the platform (memory, CPU, Input/output). An implementation for the ARINC 653 services (the APEX API). An interface to be able to configure the platform and its domain of use. Various instrumentation tools.
Initialization
Initialization of an ARINC 653 partition creates resources used by the partition. Resources creation (PROCESS, EVENT, SEMAPHORE...) is performed by calling API services named CREATE_xxxx.
Error handling
The process error handler is a preemptive process of the highest priority dedicated to handle partition exceptions. It is created by the service CREATE_ERROR_HANDLER during partition initialization.
The API allows the error handler to stop a faulty process (STOP_SELF). In that case, the RTOS scheduler will elicit the next process with the highest priority.
ARINC 653 does not specify how the scheduler should behave if the error handler does not stop a faulty process. In some (theoretical) cases, this could lead to an infinite loop between the faulty process and the error handler.
The error handler can obtain information about the source and the context of the exception.
Mode management
Each partition can be in several activation modes: COLD_START and WARM_START: Only the initialization process is executed, NORMAL: The initialization process is stopped, and the other partition processes are called by the RTOS scheduler depending on their priority, IDLE: No process is executed. However an implementation could still in theory execute a hidden process of the lowest priority, for example to start an infinite loop.
The SET_PARTITION_MODE service allows to manage these states. It can be called by any process in the partition. Entering the IDLE state is irreversible for the partition. Only an external event (such as a platform restart) can change the state to another mode when the partition is in this state.