One of my goals with this blog is to talk about flight software- what it looks like, how it is developed, and what it is like to be in the aerospace domain. I think it would be nice to see more discussion of this kind of programming, especially as space becomes available to more people through small satellites.
To do my part, I will discuss some of my thoughts on the Core Flight Executive and Core Flight Software system developed by NASA, along with the Operating system abstraction layer (OSAL) used by this project. This post will just be a brief introduction to lay the groundwork for a more in depth look at the system.
I want to be clear in that I only see a particular slice of the aerospace world- I don't work on planes or rockets or human rated systems . I have worked on a safety critical piece of software for UAV applications, and the science payload SAGE III, and I'm currently working on a small satellite science mission called ARCSTONE that will use CFE/CFS for its flight software. I say this to give some context in my background and where I'm coming from in this post.
In addition, the project has a significant amount of documentation available for each module, both as PDF files and Doxygen pages included in the source distribution.
A brief overview is that CFE consists of a set of software components that provide the main services used by flight software, whether that software is for space, UAVs, or other embedded applications. It turns out that when writing flight software there are some system one seems to need on every project, and CFE is an attempt to package these systems up for reuse across many projects, while providing additional modules that are not used in all projects but are still common. This is done in an operating system agnostic way, assisting portability.
These core components provide mechanisms for managing tables, sending packets between software components, logging system events, creating small messages to communicate the system's state to a user, allocating blocks of memory from a pool, performing performance monitoring, managing files, managing time and timestamping, and more.
In addition to these core modules, there are additional modules called apps that are part of the larger CFS system, where you can add whichever Apps you want to your project to get additional capabilities that are common but not neccesarily in every project. Some examples of these additional modules are limit monitoring, stored command execution, telemtry packet creation (HouseKeeping), file management, data storage, and task scheduling.
All of these software modules do things like spawn tasks, create and use message queues, semaphores, files, etc. When they need to do these things, they do not call operating system functions directly, as that would tie the code to a particular system. Instead they call functions defined in the header files of the OSAL project. The OSAL software provides an interface which is then implemented in a set of C files for a specific operating system. The operating systems available on github are Posix (which has worked for me on both desktop and embedded Linux), VxWorks 6.7, and RTEMS. I've only had experience running the Posix implementation of OSAL on Linux systems.
There seems to be a common confusing about the OSAL layer and CFS in general- if you write software as a CFS application there is absolutely nothing hiding the operating system or hardware from you. It is more accurate to say that you can choose to use the OSAL (and PSP) functions to make your software more portable, but if you need low level access to hardware you do so just as you might in other software.
The CFE/CFS project has an extensive makefile system which builds the core components including OSAL, CFE, and the PSP (the platform specific code that is not part of OSAL), and can be extended to call makefiles in subfolders of the apps directory in order to build additional modules.
Once CFS is built, you end up with a single executable file, a series of object files, and a cfe_es_startup.scr. When the executable is run (core-linux.bin for the POSIX build), it starts up the CFE components, and reads the cfe_es_startup.scr file. This file lists the additional modules to load, and the executable will load each one in turn and call an initialization function listed in the cfe_es_startup.scr file.
At this point you have a series of operating system threads running, message queues created, semaphores, and files open. The CFS flight software is running, logging, updating time, perhaps executing a schedule of task executions using the SCH application, creating telemetry packets which might be stored in a file by the DS app, and so on.
You have to provide input and output to this system somehow, often with a project specific module which know about your hardware's interfaces and perhaps the ground system you are talking to. There are some modules called Command Ingest and Telemetry Output, each of which has a "lab" version for development which forward commands over UDP to send to the Software Bus, and receive telemetry packets from the Software Bus to send out over UDP.
Hopefully this provides some references and a vague idea of what CFE/CFS is. I hope to keep posting about this software, and get more into the architecture and tradeoffs it makes in future posts.