Basic principles of partitioning ARINC 653

1 basic principles of partitioning

1.1 arinc 653 platform
1.2 initialization
1.3 error handling
1.4 mode management
1.5 processes of partition

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 timer , space partitioning constraints of platform (memory, cpu, input/output).
an implementation arinc 653 services (the apex api).
an interface able configure platform , domain of use.
various instrumentation tools.

initialization

initialization of arinc 653 partition creates resources used partition. resources creation (process, event, semaphore...) performed calling api services named create_xxxx.

error handling

the process error handler preemptive process of highest priority dedicated handle partition exceptions. created service create_error_handler during partition initialization.

the api allows error handler stop faulty process (stop_self). in case, rtos scheduler elicit next process highest priority.

arinc 653 not specify how scheduler should behave if error handler not stop faulty process. in (theoretical) cases, lead infinite loop between faulty process , error handler.

the error handler can obtain information source , context of exception.

mode management

each partition can in several activation modes:

cold_start , warm_start: initialization process executed,
normal: initialization process stopped, , other partition processes called rtos scheduler depending on priority,
idle: no process executed. implementation still in theory execute hidden process of lowest priority, example start infinite loop.

the set_partition_mode service allows manage these states. can called process in partition. entering idle state irreversible partition. external event (such platform restart) can change state mode when partition in state.

the processes of partition

each partition has @ least 1 process.

process scheduling preemptive. scheduler called either timer or api services.