Ogma is a tool to facilitate the integration of safe runtime monitors into other systems. Ogma extends Copilot, a high-level runtime verification framework that generates hard real-time C99 code.

• Generating NASA Core Flight System runtime monitoring applications that monitor data received from the message bus.

• Generating Robot Operating System runtime monitoring applications.

• Generating F' (FPrime) runtime monitoring components.

• Generating message handlers for NASA Core Flight System applications to make external data in structs available to a Copilot monitor.

• Generating the glue code necessary to work with C structs in Copilot.

Integration of monitors into larger applications (e.g., simulators).

• Installation
  • Pre-requisites
  • Compilation
• Usage
  • cFS Application Generation
  • ROS Application Generation
  • F' Component Generation
  • Struct Interface Generation
• Contributions
• Acknowledgements
• License

^{(Back to top)}

^{(Back to top)}

To install Ogma from source, users must have the tools GHC and cabal-install. At this time, we recommend GHC 8.6 and a version of cabal-install between 2.4 and 3.2. (Ogma has been tested with GHC versions up to 9.2 and cabal-install versions up to 3.6, although the installation steps may vary slightly depending on the version of cabal-install being used.)

On Debian or Ubuntu Linux, both can be installed with:

$ apt-get install ghc cabal-install

On Mac, they can be installed with:

$ brew install ghc cabal-install

^{(Back to top)}

Once GHC and cabal are installed, the simplest way to install Ogma is with:

$ git clone https://github.com/nasa/ogma.git
$ cd ogma
$ export PATH="$HOME/.cabal/bin/:$PATH"
$ cabal v1-update
$ cabal v1-install alex happy
$ cabal v1-install BNFC copilot
$ cabal v1-install ogma-*/

After that, the ogma executable will be placed in the directory $HOME/.cabal/bin/, where $HOME represents your user's home directory.

^{(Back to top)}

The main invocation of ogma with --help lists sub-commands available. More details are provided in the following sections.

NASA Core Flight System (cFS) is a flight software architecture to implement complex systems by combining multiple reusable applications that communicate to one another via a software bus. cFS has been used, among others, on spacecraft, cubesats, and drones.

Ogma includes multiple facilities to generate cFS applications. The cFS applications generated by Ogma perform three steps to connect Copilot monitors to the application:

• Subscribe to a message in the cFS communication bus.
• When a message of the desired kind arrives, copy the data to make it available to Copilot and call the monitor's main entry point.
• Declare handlers that are executed when the property being monitored is violated.

When using this facility, Ogma produces a Copilot file that the user is expected to modify to implement the property to monitor. To avoid having to modify the generated C files that implement the cFS app itself, Ogma gives the ability to state what information one is interested in monitoring. If the kind of information is known to Ogma, it will automatically subscribe to the necessary messages and make it available to Copilot. Ogma provides additional flags to customize the list of known variables, so that projects can maintain their own variable databases beyond what Ogma includes by default.

cFS applications are generated using the Ogma command cfs, which receives four main arguments:

• --app-target-dir DIR: location where the cFS application files must be stored.
• --app-template-dir DIR: location of the cFS application template to use.
• --variable-file FILENAME: a file containing a list of variables that must be made available to the monitor.
• --variable-db FILENAME: a file containing a database of known variables, and the message they are included with.

The following execution generates an initial cFS application for runtime monitoring using Copilot:

$ ogma cfs --variable-db examples/cfs-variable-db --variable-file examples/cfs-variables

The application generated by Ogma contains the following files:

copilot-cfs-demo/CMakeLists.txt
copilot-cfs-demo/fsw/for_build/Makefile
copilot-cfs-demo/fsw/mission_inc/copilot_cfs_perfids.h
copilot-cfs-demo/fsw/platform_inc/copilot_cfs_msgids.h
copilot-cfs-demo/fsw/src/copilot_cfs.c
copilot-cfs-demo/fsw/src/Properties.hs
copilot-cfs-demo/fsw/src/copilot_cfs_msg.h
copilot-cfs-demo/fsw/src/copilot_cfs_events.h
copilot-cfs-demo/fsw/src/copilot_cfs_version.h
copilot-cfs-demo/fsw/src/copilot_cfs.h

Users are expected to modify Properties.hs to adjust the property being monitored. Although it is possible to adjust the file copilot_cfs.c to include property violation handlers, we recommend adding them in a separate C file and modifying the compilation scripts to include that additional file. That way, invoking Ogma again will not overwrite the changes made to the cFS application.

In this particular example, the C code generated contains the following instruction to subscribe to an ICAROUS_POSITION_MID message to obtain the vehicle position:

    CFE_SB_Subscribe(ICAROUS_POSITION_MID, COPILOT_CommandPipe);

The message dispatcher included in the application detects a message of this kind and calls a dedicated subroutine:

void COPILOT_ProcessCommandPacket(void)
{
    CFE_SB_MsgId_t  MsgId;

    MsgId = CFE_SB_GetMsgId(COPILOTMsgPtr);

    switch (MsgId)
    {
        case ICAROUS_POSITION_MID:
            COPILOT_ProcessIcarousPosition();
            break;
    ...

Finally, the dedicated subroutine makes data available to the monitor and calls the main Copilot entry point step:

void COPILOT_ProcessIcarousPosition(void)
{
    position_t* msg;
    msg = (position_t*) COPILOTMsgPtr;
    position = *msg;
    step();
}

By default, Ogma uses a pre-defined template to generate the cFS monitoring application. It's possible to customize the output by providing a directory with a set of files with a cFS application template, which Ogma will use instead.

To choose this feature, one must call Ogma's cfs command with the argument --app-template-dir DIR, where DIR is the path to a directory containing a cFS application template. For example, assuming that the directory my_template contains a custom cFS application template, one can execute:

$ ogma cfs --app-template-dir my_template/ --variable-db examples/cfs-variable-db --variable-file examples/cfs-variables

Ogma will copy the files in that directory to the target path, filling in several holes with specific information:

• {{variablesS}}: this will be replaced by a list of variable declarations, one for each global variable that holds information read from the cFS software bus that must be made accessible to the monitoring code.

• {{msgSubscriptionsS}}: this will be replaced by a list of calls to CFE_SB_Subscribe, subscribing to the necessary information coming in the software bus.

• {{msgCasesS}}: this will be replaced by a switch case statements that match the ID of an incoming message, to handle information being received that must be updated and would trigger a re-evaluation of the monitors.

• {{msgHandlerS}}: this will be replaced by function definitions of the functions that will be called to actually update the variables with information coming from the software bus, and re-evaluate the monitors.

We understand that this level of customization may be insufficient for your application. If that is the case, feel free to reach out to our team to discuss how we could make the template expansion system more versatile.

The Robot Operating System (ROS) is a framework to build robot applications.

Ogma is able to generate ROS monitoring applications that subscribe to obtain the data needed by the monitors and report any violations. At present, support for ROS app generation is considered preliminary.

ROS applications are generated using the Ogma command ros, which receives four main arguments:

• --app-target-dir DIR: location where the ROS application files must be stored.
• --variable-file FILENAME: a file containing a list of variables that must be made available to the monitor.
• --variable-db FILENAME: a file containing a database of known variables, and the topic they are included with.
• --handlers FILENAME: a file containing a list of handlers used in the specification.

The following execution generates an initial ROS application for runtime monitoring using Copilot:

$ ogma ros --handlers filename --variable-file variables --variable-db ros-variable-db --app-target-dir ros_demo

The application generated by Ogma contains the following files:

ros_demo/copilot/CMakeLists.txt
ros_demo/copilot/src/copilot_monitor.cpp
ros_demo/copilot/src/copilot_logger.cpp
ros_demo/copilot/src/.keep
ros_demo/copilot/package.xml
ros_demo/Dockerfile

The Dockerfile can be used to compile the application inside the base image for the Space ROS distribution. To build the image, first place the core implementation of the monitors in C inside the directory ros_demo/copilot/src/ (see 'Current limitations' for details). After, you can compile the image with:

docker build .

The argument variable DB passed to the ROS backend should contain a list of variables, together with their types and the corresponding ROS topic in which those variables are passed. Each line in that file has the format:

("<NAME>","<TYPE>","<TOPIC_NAME>","<RESERVED_FOR_FUTURE_USE>")

For example, an input variable called "temperature" of type 64-bit signed integer coming in a ROS topic called "/battery/temperature" should have a matching entry in the variable DB file like the following:

("temperature","int64_t","/battery/temperature","ignore")

There should never be two lines in the same file with the same variable name. Variables in the DB that are not not used in any of the properties being monitored and/or are not listed in the variable file passed as argument to the ROS command will be ignored.

For a more concrete example, see the files in ogma-cli/examples/ros-copilot/ and the last step of the script .github/workflows/repo-ghc-8.6-cabal-2.4-ros.yml, which generates a ROS monitor with multiple variables and compiles the resulting code.

The user must place the code generated by Copilot monitors in two files, ros_demo/src/monitor.h and ros_demo/src/monitor.c. No Copilot or C code for the monitors is generated by default.

The code generated by default assumes that handlers receive no arguments. The user must modify the handlers accordingly if that is not the case.

Although the variable DB file is not mandatory, it is in practice required to monitor any requirement that uses any input data: no topic subscriptions will be generated for any variables for which a DB entry cannot be found. At present, Ogma will proceed without warnings if a variable is mentioned in a requirement or variables file but a matching entry is not found in the variable DB.

F' (FPrime) is a component-based framework for spaceflight applications.

Ogma is able to generate F' monitoring components that subscribe to obtain the data needed by the monitors and report any violations. At present, support for F' component generation is considered preliminary.

F' components are generated using the Ogma command fprime, which receives five main arguments:

• --app-target-dir DIR: location where the F' application files must be stored.
• --variable-file FILENAME: a file containing a list of variables that must be made available to the monitor.
• --variable-db FILENAME: a file containing a database of known variables, and their types.
• --handlers FILENAME: a file containing a list of handlers used in the specification.

The following execution generates an initial F' component for runtime monitoring using Copilot:

$ ogma fprime --handlers filename --variable-file filename --variable-db fprime-variable-db --app-target-dir fprime_demo

The component generated by Ogma contains the following files:

fprime_demo/CMakeLists.txt
fprime_demo/Copilot.fpp
fprime_demo/Copilot.cpp
fprime_demo/Copilot.hpp
fprime_demo/Dockerfile
fprime_demo/inline-copilot

The format of the variables, variable DB, and handlers file are as follows.

The variables file can contain a list of variables used in a specification, one per line. For example, if we are working with a specification that uses three boolean variables called autopilot, sensorLimitsExceeded, and pullup, we can provide them to Ogma's fprime command in a file like the following:

$ cat variables
autopilot
sensorLimitsExceeded
pullup

The variables database file contains a list of known variables and their types. It does not matter if there are variables that are not used for one particular specification, or property/requirement/monitor. The only thing that matters is that the variables used, and their types, be listed in the file. Continuing with the same example, we could have:

$ cat fprime-variable-db
("temperature", "uint8_t")
("autopilot", "bool")
("sensorLimitsExceeded", "bool")
("pullup", "bool")
("current_consumption", "float")

In our example, we only care about the boolean variables; it is sufficient that they be listed in the variable DB file.

Finally, the handlers file is a list of monitor handlers that the generated FPrime component should restrict to monitoring. They are listed one per line:

$ cat handlers
handlerpropREQ_001

The user must place the code generated by Copilot monitors in three files, fprime_demo/src/copilot.h, fprime_demo/src/copilot_types.h and fprime_demo/src/copilot.c. No Copilot or C code for the monitors is generated by default by the fprime command.

The code generated by default assumes that handlers receive no arguments. The user must modify the handlers accordingly if that is not the case.

A lot of the information that must be monitored in real-world C applications is packed in structs. Copilot allows accessing specific fields of C structs, but requires additional definitions in the Copilot language to make the shape of those structs known to the compiler.

Ogma is able to generate the boilerplate code needed to work with C structs in Copilot. For example, to use the following struct as the type of an extern stream in Copilot, the user is expected to define several Copilot (Haskell) types and type class instances:

typedef struct {
   double x;
   double y;
} point;

Ogma can generate that code automatically with the structs subcommand:

$ ogma structs --header-file-name examples/point.h
data Point = Point
  { pX :: Field "x" Double
  , pY :: Field "y" Double
  }

instance Struct Point where
  typename _ = "point"
  toValues v = [ Value Double (pX v), Value Double (pY v) ]

instance Typed Point where
  typeOf = Struct (Point (Field 0) (Field 0))

By including these definitions in a Copilot file, users can now access the individual x and y fields of a Point in a stream.

^{(Back to top)}

The best way to contribute to Ogma is to report any issues you find via the issue tracker, and to use Ogma to build applications or in your own research and let us know about your results.

We kindly ask users not to send PRs to this project. Instead, please document the bugs you find or other suggestions as issues and we will make the necessary changes.

^{(Back to top)}

Ogma has been created by Ivan Perez and Alwyn Goodloe.

The Ogma team would like to thank Swee Balachandran, Dimitra Giannakopoulou, Anastasia Mavridou, Cesar Munoz and Thomas Pressburger, for the input during the development of Ogma.

X-Plane images obtained via the X-Plane 10 (Pro) flight simulator. Re-shared with permission.

^{(Back to top)}

Copyright 2020-2021 United States Government as represented by the Administrator of the National Aeronautics and Space Administration. All Rights Reserved.

See the file LICENSE.pdf for details.