KSat/fsfw
15
4
Fork 0
forked from fsfw/fsfw
Code Issues Pull Requests 1 Releases Wiki Activity

Merge branch 'development' into mueller/action-bugfix

Browse Source
This commit is contained in:
Steffen Gaisser 2021-01-19 14:52:58 +01:00
commit 2469da757e
13 changed files with 228 additions and 164 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
CMakeLists.txt
View File

@ -106,7 +106,7 @@ else()
)
endif()
if(CMAKE_COMPILER_IS_GNUCXX)
if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
set(WARNING_FLAGS
-Wall
-Wextra

133
README.md
View File

@ -9,121 +9,40 @@ The initial version of the Flight Software Framework was developed during
the Flying Laptop Project by the University of Stuttgart in cooperation
with Airbus Defence and Space GmbH.
## Intended Use
## Quick facts
The framework is designed for systems, which communicate with external devices, perform control loops, receive telecommands and send telemetry, and need to maintain a high level of availability.
Therefore, a mode and health system provides control over the states of the software and the controlled devices.
In addition, a simple mechanism of event based fault detection, isolation and recovery is implemented as well.
The framework is designed for systems, which communicate with external devices, perform control loops, receive telecommands and send telemetry, and need to maintain a high level of availability. Therefore, a mode and health system provides control over the states of the software and the controlled devices. In addition, a simple mechanism of event based fault detection, isolation and recovery is implemented as well.
The recommended hardware is a microprocessor with more than 1 MB of RAM and 1 MB of non-volatile Memory.
For reference, current Applications use a Cobham Gaisler UT699 (LEON3FT), a ISISPACE IOBC or a Zynq-7020 SoC.
The `fsfw` was also tested on the STM32H743ZI-Nucleo board.
The FSFW provides abstraction layers for operating systems to provide a uniform operating system abstraction layer (OSAL). Some components of this OSAL are required internally by the FSFW but is also very useful for developers to implement the same application logic on different operating systems with a uniform interface.
## How to Use
Currently, the FSFW provides the following OSALs:
- Linux
- Host
- FreeRTOS
- RTEMS
The recommended hardware is a microprocessor with more than 1 MB of RAM and 1 MB of non-volatile Memory. For reference, current applications use a Cobham Gaisler UT699 (LEON3FT), a ISISPACE IOBC or a Zynq-7020 SoC. The `fsfw` was also successfully run on the STM32H743ZI-Nucleo board and on a Raspberry Pi and is currently running on the active satellite mission Flying Laptop.
## Getting started
The [FSFW example](https://egit.irs.uni-stuttgart.de/fsfw/fsfw_example) provides a good starting point and a demo to see the FSFW capabilities and build it with the Make or the CMake build system. It is recommended to evaluate the FSFW by building and playing around with the demo application.
The [FSFW example](https://egit.irs.uni-stuttgart.de/fsfw/fsfw_example) provides a good starting point and a demo
to see the FSFW capabilities and build it with the Make or the CMake build system.
Generally, the FSFW is included in a project by compiling the FSFW sources and providing
a configuration folder and adding it to the include path.
a configuration folder and adding it to the include path. There are some functions like `printChar` which are different depending on the target architecture and need to be implemented by the mission developer.
A template configuration folder was provided and can be copied into the project root to have
a starting point. The [configuration section](doc/README-config.md#top) provides more specific information about
the possible options.
a starting point. The [configuration section](doc/README-config.md#top) provides more specific information about the possible options.
## Structure
## Index
The general structure is driven by the usage of interfaces provided by objects.
The FSFW uses C++11 as baseline. The intention behind this is that this C++ Standard should be widely available, even with older compilers.
The FSFW uses dynamic allocation during the initialization but provides static containers during runtime.
This simplifies the instantiation of objects and allows the usage of some standard containers.
Dynamic Allocation after initialization is discouraged and different solutions are provided in the FSFW to achieve that.
The fsfw uses run-time type information but exceptions are not allowed.
[1. High-level overview](doc/README-highlevel.md#top) <br>
[2. Core components](doc/README-core.md#top) <br>
[3. OSAL overview](doc/README-osal.md#top) <br>
[4. PUS services](doc/README-pus.md#top) <br>
[5. Device Handler overview](doc/README-devicehandlers.md#top) <br>
[6. Controller overview](doc/README-controllers.md#top) <br>
[7. Local Data Pools](doc/README-localpools.md#top) <br>
### Failure Handling
Functions should return a defined ReturnValue_t to signal to the caller that something has gone wrong.
Returnvalues must be unique. For this the function HasReturnvaluesIF::makeReturnCode or the Macro MAKE_RETURN can be used.
The CLASS_ID is a unique id for that type of object. See returnvalues/FwClassIds.
### OSAL
The FSFW provides operation system abstraction layers for Linux, FreeRTOS and RTEMS.
A independent Host OSAL is in development which will provide abstraction for common type of
host OSes (tested for Linux and Windows, not for MacOS yet).
The OSAL provides periodic tasks, message queues, clocks and semaphores as well as mutexes.
### Core Components
The FSFW has following core components. More detailed informations can be found in the
[core component section](doc/README-core.md#top):
1. Tasks: Abstraction for different (periodic) task types like periodic tasks or tasks with fixed timeslots
2. ObjectManager: This module stores all `SystemObjects` by mapping a provided unique object ID to the object handles.
3. Static Stores: Different stores are provided to store data of variable size (like telecommands or small telemetry) in a pool structure without
using dynamic memory allocation. These pools are allocated up front.
3. Clock: This module provided common time related functions
4. EventManager: This module allows routing of events generated by `SystemObjects`
5. HealthTable: A component which stores the health states of objects
### Static Ids in the framework
Some parts of the framework use a static routing address for communication.
An example setup of ids can be found in the example config in "defaultcft/fsfwconfig/objects/Factory::setStaticFrameworkObjectIds()".
### Events
Events are tied to objects. EventIds can be generated by calling the Macro MAKE_EVENT. This works analog to the returnvalues.
Every object that needs own EventIds has to get a unique SUBSYSTEM_ID.
Every SystemObject can call triggerEvent from the parent class.
Therefore, event messages contain the specific EventId and the objectId of the object that has triggered.
### Internal Communication
Components communicate mostly over Message through Queues.
Those queues are created by calling the singleton QueueFactory::instance()->create().
### External Communication
The external communication with the mission control system is mostly up to the user implementation.
The FSFW provides PUS Services which can be used to but don't need to be used.
The services can be seen as a conversion from a TC to a message based communication and back.
#### CCSDS Frames, CCSDS Space Packets and PUS
If the communication is based on CCSDS Frames and Space Packets, several classes can be used to distributed the packets to the corresponding services. Those can be found in tcdistribution.
If Space Packets are used, a timestamper must be created.
An example can be found in the timemanager folder, this uses CCSDSTime::CDS_short.
#### Device Handlers
DeviceHandlers are another important component of the FSFW.
The idea is, to have a software counterpart of every physical device to provide a simple mode, health and commanding interface.
By separating the underlying Communication Interface with DeviceCommunicationIF, a device handler (DH) can be tested on different hardware.
The DH has mechanisms to monitor the communication with the physical device which allow for FDIR reaction.
Device Handlers can be created by overriding `DeviceHandlerBase`.
A standard FDIR component for the DH will be created automatically but can be overwritten by the user.
More information on DeviceHandlers can be found in the related [documentation section](doc/README-devicehandlers.md#top).
#### Modes, Health
The two interfaces HasModesIF and HasHealthIF provide access for commanding and monitoring of components.
On-board Mode Management is implement in hierarchy system.
DeviceHandlers and Controllers are the lowest part of the hierarchy.
The next layer are Assemblies. Those assemblies act as a component which handle redundancies of handlers.
Assemblies share a common core with the next level which are the Subsystems.
Those Assemblies are intended to act as auto-generated components from a database which describes the subsystem modes.
The definitions contain transition and target tables which contain the DH, Assembly and Controller Modes to be commanded.
Transition tables contain as many steps as needed to reach the mode from any other mode, e.g. a switch into any higher AOCS mode might first turn on the sensors, than the actuators and the controller as last component.
The target table is used to describe the state that is checked continuously by the subsystem.
All of this allows System Modes to be generated as Subsystem object as well from the same database.
This System contains list of subsystem modes in the transition and target tables.
Therefore, it allows a modular system to create system modes and easy commanding of those, because only the highest components must be commanded.
The health state represents if the component is able to perform its tasks.
This can be used to signal the system to avoid using this component instead of a redundant one.
The on-board FDIR uses the health state for isolation and recovery.
## Unit Tests
Unit Tests are provided in the unittest folder. Those use the catch2 framework but do not include catch2 itself.
See README.md in the unittest Folder.

42
datapool/PoolEntry.cpp
View File

@ -1,32 +1,24 @@
#include "PoolEntry.h"
#include "../serviceinterface/ServiceInterfaceStream.h"
#include "../serviceinterface/ServiceInterface.h"
#include "../globalfunctions/arrayprinter.h"
#include <cstring>
#include <algorithm>
template <typename T>
PoolEntry<T>::PoolEntry(std::initializer_list<T> initValue, uint8_t setLength,
bool setValid ) : length(setLength), valid(setValid) {
PoolEntry<T>::PoolEntry(std::initializer_list<T> initValue, bool setValid ):
length(initValue.size()), valid(setValid) {
this->address = new T[this->length];
if(initValue.size() == 0) {
std::memset(this->address, 0, this->getByteSize());
}
else if (initValue.size() != setLength){
#if FSFW_CPP_OSTREAM_ENABLED == 1
sif::warning << "PoolEntry: setLength is not equal to initializer list"
"length! Performing zero initialization with given setLength"
<< std::endl;
#endif
std::memset(this->address, 0, this->getByteSize());
}
else {
std::copy(initValue.begin(), initValue.end(), this->address);
}
}
template <typename T>
PoolEntry<T>::PoolEntry( T* initValue, uint8_t setLength, bool setValid ) :
PoolEntry<T>::PoolEntry(T* initValue, uint8_t setLength, bool setValid):
length(setLength), valid(setValid) {
this->address = new T[this->length];
if (initValue != nullptr) {
@ -70,14 +62,26 @@ bool PoolEntry<T>::getValid() {
template <typename T>
void PoolEntry<T>::print() {
const char* validString = nullptr;
if(valid) {
validString = "Valid";
}
else {
validString = "Invalid";
}
#if FSFW_CPP_OSTREAM_ENABLED == 1
sif::debug << "Pool Entry Validity: " <<
(this->valid? " (valid) " : " (invalid) ") << std::endl;
#endif
arrayprinter::print(reinterpret_cast<uint8_t*>(address), length);
#if FSFW_CPP_OSTREAM_ENABLED == 1
sif::debug << std::dec << std::endl;
sif::info << "PoolEntry information." << std::endl;
sif::info << "PoolEntry validity: " << validString << std::endl;
#else
sif::printInfo("PoolEntry information.\n");
sif::printInfo("PoolEntry validity: %s\n", validString);
#endif
arrayprinter::print(reinterpret_cast<uint8_t*>(address), getByteSize());
}
template<typename T>
inline T* PoolEntry<T>::getDataPtr() {
return this->address;
}
template<typename T>
@ -88,8 +92,10 @@ Type PoolEntry<T>::getType() {
template class PoolEntry<uint8_t>;
template class PoolEntry<uint16_t>;
template class PoolEntry<uint32_t>;
template class PoolEntry<uint64_t>;
template class PoolEntry<int8_t>;
template class PoolEntry<int16_t>;
template class PoolEntry<int32_t>;
template class PoolEntry<int64_t>;
template class PoolEntry<float>;
template class PoolEntry<double>;

61
datapool/PoolEntry.h
View File

@ -35,24 +35,22 @@ public:
"uint8_t");
/**
* @brief In the classe's constructor, space is allocated on the heap and
* potential init values are copied to that space.
* potential initialization values are copied to that space.
* @details
* Not passing any arguments will initialize an non-array pool entry
* (setLength = 1) with an initial invalid state.
* Please note that if an initializer list is passed, the correct
* corresponding length should be passed too, otherwise a zero
* initialization will be performed with the given setLength.
* with an initial invalid state and the value 0.
* Please note that if an initializer list is passed, the length of the
* initializer list needs to be correct for vector entries because
* required allocated space will be deduced from the initializer list length
* and the pool entry type.
* @param initValue
* Initializer list with values to initialize with, for example {0,0} to
* initialize the two entries to zero.
* @param setLength
* Defines the array length of this entry. Should be equal to the
* intializer list length.
* Initializer list with values to initialize with, for example {0, 0} to
* initialize the a pool entry of a vector with two entries to 0.
* @param setValid
* Sets the initialization flag. It is invalid by default.
*/
PoolEntry(std::initializer_list<T> initValue = {}, uint8_t setLength = 1,
bool setValid = false);
PoolEntry(std::initializer_list<T> initValue = {0}, bool setValid = false);
/**
* @brief In the classe's constructor, space is allocated on the heap and
* potential init values are copied to that space.
@ -66,9 +64,9 @@ public:
*/
PoolEntry(T* initValue, uint8_t setLength = 1, bool setValid = false);
//! Explicitely deleted copy ctor, copying is not allowed!
//! Explicitely deleted copy ctor, copying is not allowed.
PoolEntry(const PoolEntry&) = delete;
//! Explicitely deleted copy assignment, copying is not allowed!
//! Explicitely deleted copy assignment, copying is not allowed.
PoolEntry& operator=(const PoolEntry&) = delete;
/**
@ -82,21 +80,16 @@ public:
~PoolEntry();
/**
* @brief This is the address pointing to the allocated memory.
* Return typed pointer to start of data.
* @return
*/
T* address;
/**
* @brief This attribute stores the length information.
*/
uint8_t length;
/**
* @brief Here, the validity information for a variable is stored.
* Every entry (single variable or vector) has one valid flag.
*/
uint8_t valid;
T* getDataPtr();
/**
* @brief getSize returns the array size of the entry.
* @details A single parameter has size 1.
* @details
* For non-array pool entries return type size, for vector entries
* return type size times the number of entries.
*/
uint8_t getSize();
/**
@ -123,8 +116,22 @@ public:
* information to the screen. It prints all array entries in a row.
*/
void print();
Type getType();
private:
/**
* @brief This attribute stores the length information.
*/
uint8_t length;
/**
* @brief Here, the validity information for a variable is stored.
* Every entry (single variable or vector) has one valid flag.
*/
uint8_t valid;
/**
* @brief This is the address pointing to the allocated memory.
*/
T* address;
};
#endif /* FSFW_DATAPOOL_POOLENTRY_H_ */

10
defaultcfg/fsfwconfig/FSFWConfig.h
View File

@ -9,18 +9,14 @@
//! the C stdio functions can be used alternatively
#define FSFW_CPP_OSTREAM_ENABLED 1
//! More FSFW related printouts.
//! Be careful, this also turns off most diagnostic prinouts!
#define FSFW_ENHANCED_PRINTOUT 0
//! More FSFW related printouts depending on level. Useful for development.
#define FSFW_VERBOSE_LEVEL 1
//! Can be used to completely disable printouts, even the C stdio ones.
#if FSFW_CPP_OSTREAM_ENABLED == 0 && FSFW_ENHANCED_PRINTOUT == 0
#if FSFW_CPP_OSTREAM_ENABLED == 0 && FSFW_VERBOSE_LEVEL == 0
#define FSFW_DISABLE_PRINTOUT 0
#endif
//! Can be used to enable additional debugging printouts for developing the FSFW
#define FSFW_PRINT_VERBOSITY_LEVEL 0
//! Can be used to disable the ANSI color sequences for C stdio.
#define FSFW_COLORED_OUTPUT 1

1
doc/README-controllers.md Normal file
View File

@ -0,0 +1 @@
## Controllers

5
doc/README-core.md
View File

@ -16,9 +16,8 @@ Flight Software Framework
* A reference to an object can be get by calling the following function. T must be the specific Interface you want to call.
A nullptr check of the returning Pointer must be done. This function is based on Run-time type information.
``` c++
template <typename T> T* ObjectManagerIF::get( object_id_t id )
```cpp
template <typename T> T* ObjectManagerIF::get( object_id_t id )
```
* A typical way to create all objects on startup is a handing a static produce function to the ObjectManager on creation.
By calling objectManager->initialize() the produce function will be called and all SystemObjects will be initialized afterwards.

2
doc/README-devicehandlers.md
View File

@ -1 +1 @@
## FSFW DeviceHandlers
## Device Handlers

99
doc/README-highlevel.md Normal file
View File

@ -0,0 +1,99 @@
# High-level overview
## Structure
The general structure is driven by the usage of interfaces provided by objects.
The FSFW uses C++11 as baseline. The intention behind this is that this C++ Standard should be widely available, even with older compilers.
The FSFW uses dynamic allocation during the initialization but provides static containers during runtime.
This simplifies the instantiation of objects and allows the usage of some standard containers.
Dynamic Allocation after initialization is discouraged and different solutions are provided in the FSFW to achieve that.
The fsfw uses run-time type information but exceptions are not allowed.
### Failure Handling
Functions should return a defined ReturnValue_t to signal to the caller that something has gone wrong.
Returnvalues must be unique. For this the function HasReturnvaluesIF::makeReturnCode or the Macro MAKE_RETURN can be used.
The CLASS_ID is a unique id for that type of object. See returnvalues/FwClassIds.
### OSAL
The FSFW provides operation system abstraction layers for Linux, FreeRTOS and RTEMS.
The OSAL provides periodic tasks, message queues, clocks and semaphores as well as mutexes.
The [OSAL README](doc/README-osal.md#top) provides more detailed information on provided components and how to use them.
### Core Components
The FSFW has following core components. More detailed informations can be found in the
[core component section](doc/README-core.md#top):
1. Tasks: Abstraction for different (periodic) task types like periodic tasks or tasks with fixed timeslots
2. ObjectManager: This module stores all `SystemObjects` by mapping a provided unique object ID to the object handles.
3. Static Stores: Different stores are provided to store data of variable size (like telecommands or small telemetry) in a pool structure without
using dynamic memory allocation. These pools are allocated up front.
3. Clock: This module provided common time related functions
4. EventManager: This module allows routing of events generated by `SystemObjects`
5. HealthTable: A component which stores the health states of objects
### Static IDs in the framework
Some parts of the framework use a static routing address for communication.
An example setup of ids can be found in the example config in "defaultcft/fsfwconfig/objects/Factory::setStaticFrameworkObjectIds()".
### Events
Events are tied to objects. EventIds can be generated by calling the Macro MAKE_EVENT. This works analog to the returnvalues.
Every object that needs own EventIds has to get a unique SUBSYSTEM_ID.
Every SystemObject can call triggerEvent from the parent class.
Therefore, event messages contain the specific EventId and the objectId of the object that has triggered.
### Internal Communication
Components communicate mostly over Message through Queues.
Those queues are created by calling the singleton QueueFactory::instance()->create().
### External Communication
The external communication with the mission control system is mostly up to the user implementation.
The FSFW provides PUS Services which can be used to but don't need to be used.
The services can be seen as a conversion from a TC to a message based communication and back.
#### CCSDS Frames, CCSDS Space Packets and PUS
If the communication is based on CCSDS Frames and Space Packets, several classes can be used to distributed the packets to the corresponding services. Those can be found in tcdistribution.
If Space Packets are used, a timestamper must be created.
An example can be found in the timemanager folder, this uses CCSDSTime::CDS_short.
#### Device Handlers
DeviceHandlers are another important component of the FSFW.
The idea is, to have a software counterpart of every physical device to provide a simple mode, health and commanding interface.
By separating the underlying Communication Interface with DeviceCommunicationIF, a device handler (DH) can be tested on different hardware.
The DH has mechanisms to monitor the communication with the physical device which allow for FDIR reaction.
Device Handlers can be created by overriding `DeviceHandlerBase`.
A standard FDIR component for the DH will be created automatically but can be overwritten by the user.
More information on DeviceHandlers can be found in the related [documentation section](doc/README-devicehandlers.md#top).
#### Modes, Health
The two interfaces HasModesIF and HasHealthIF provide access for commanding and monitoring of components.
On-board Mode Management is implement in hierarchy system.
DeviceHandlers and Controllers are the lowest part of the hierarchy.
The next layer are Assemblies. Those assemblies act as a component which handle redundancies of handlers.
Assemblies share a common core with the next level which are the Subsystems.
Those Assemblies are intended to act as auto-generated components from a database which describes the subsystem modes.
The definitions contain transition and target tables which contain the DH, Assembly and Controller Modes to be commanded.
Transition tables contain as many steps as needed to reach the mode from any other mode, e.g. a switch into any higher AOCS mode might first turn on the sensors, than the actuators and the controller as last component.
The target table is used to describe the state that is checked continuously by the subsystem.
All of this allows System Modes to be generated as Subsystem object as well from the same database.
This System contains list of subsystem modes in the transition and target tables.
Therefore, it allows a modular system to create system modes and easy commanding of those, because only the highest components must be commanded.
The health state represents if the component is able to perform its tasks.
This can be used to signal the system to avoid using this component instead of a redundant one.
The on-board FDIR uses the health state for isolation and recovery.
## Unit Tests
Unit Tests are provided in the unittest folder. Those use the catch2 framework but do not include catch2 itself. More information on how to run these tests can be found in the separate
[`fsfw_tests` reposoitory](https://egit.irs.uni-stuttgart.de/fsfw/fsfw_tests)

3
doc/README-localpools.md Normal file
View File

@ -0,0 +1,3 @@
## Local Data Pools

32
doc/README-osal.md Normal file
View File

@ -0,0 +1,32 @@
# Operating System Abstraction Layer (OSAL)
Some specific information on the provided OSALs are provided.
## Linux OSAL
This OSAL can be used to compile for Linux host systems like Ubuntu 20.04 or for
embedded Linux targets like the Raspberry Pi. This OSAL generally requires threading support
and real-time functionalities. For most UNIX systems, this is done by adding `-lrt` and `-lpthread` to the linked libraries in the compilation process. The CMake build support provided will do this automatically for the `fsfw` target. It should be noted that most UNIX systems need to be configured specifically to allow the real-time functionalities required by the FSFW.
More information on how to set up a Linux system accordingly can be found in the
[Linux README of the FSFW example application](https://egit.irs.uni-stuttgart.de/fsfw/fsfw_example/src/branch/master/doc/README-linux.md#top)
## Hosted OSAL
This is the newest OSAL. Support for Semaphores has not been implemented yet and will propably be implemented as soon as C++20 with Semaphore support has matured. This OSAL can be used to run the FSFW on any host system, but currently has only been tested on Windows 10 and Ubuntu 20.04. Unlike the other OSALs, it uses dynamic memory allocation (e.g. for the message queue implementation). Cross-platform serial port (USB) support might be added soon.
## FreeRTOS OSAL
FreeRTOS is not included and the developer needs to take care of compiling the FreeRTOS sources and adding the `FreeRTOSConfig.h` file location to the include path. This OSAL has only been tested extensively with the pre-emptive scheduler configuration so far but it should in principle also be possible to use a cooperative scheduler. It is recommended to use the `heap_4` allocation scheme. When using newlib (nano), it is also recommended to add `#define configUSE_NEWLIB_REENTRANT` to the FreeRTOS configuration file to ensure thread-safety.
When using this OSAL, developers also need to provide an implementation for the `vRequestContextSwitchFromISR` function. This has been done because the call to request a context switch from an ISR is generally located in the `portmacro.h` header and is different depending on the target architecture or device.
## RTEMS OSAL
The RTEMS OSAL was the first implemented OSAL which is also used on the active satellite Flying Laptop.
## TCP/IP socket abstraction
The Linux and Host OSAL provide abstraction layers for the socket API. Currently, only UDP sockets have been imlemented. This is very useful to test TMTC handling either on the host computer directly (targeting localhost with a TMTC application) or on embedded Linux devices, sending TMTC packets via Ethernet.

1
doc/README-pus.md Normal file
View File

@ -0,0 +1 @@
## PUS Services

1
pus/CMakeLists.txt
View File

@ -5,6 +5,7 @@ target_sources(${LIB_FSFW_NAME}
Service17Test.cpp
Service1TelecommandVerification.cpp
Service2DeviceAccess.cpp
Service3Housekeeping.cpp
Service5EventReporting.cpp
Service8FunctionManagement.cpp
Service9TimeManagement.cpp