Clone this wiki locally
Table of contents
RIOT is an operating system designed for the particular requirements of Internet of Things (IoT) scenarios. These requirements comprise a low memory footprint, high energy efficiency, real-time capabilities, a modular and configurable communication stack, and support for a wide range of low-power devices. RIOT provides a microkernel, utilities like cryptographic libraries, data structures (bloom filters, hash tables, priority queues), or a shell, different network stacks, and support for various microcontrollers, radio drivers, sensors, and configurations for entire platforms, e.g. TelosB or STM32 Discovery Boards.
The microkernel itself comprises thread management, a priority-based scheduler, a powerful API for inter-process communication (IPC), a system timer, and mutexes.
In order to build an application or library with RIOT, you need first to
download the source code (Getting the source
code). This contains - besides the
before mentioned features - also some example applications (located in the
examples subdirectory) and a sample Makefile you may use for your own
application. This Makefile template shows you how to compile and link your application
against RIOT (Compiling RIOT).
Native RIOT - Run RIOT on your PC!
As a special platform, you will find a CPU and board called
native in the
repository. This target allows you to run RIOT as a process on Linux on most
supported hardware platforms. Just set BOARD to
native in your
application's Makefile, call
make, and execute the resulting elf-file. Further
documentation about the native port can be found in
The RIOT repository contains the following ten subdirectories:
boards directory provides the configurations and initialization code for
supported IoT platforms. In
core you can find the kernel, while
comprises microcontroller specific implementations like startup and exception
handling code. The folder
dist contains a template for an application's Makefile
and external utilities like the terminal program
pyterm or a script to build
your own toolchain for ARM microcontrollers. Not very surprisingly you will find
the (doxygen) documentation in
doc and peripheral driver code in
examples folder provides some exemplary applications,
Makefiles to integrate external libraries into RIOT, and
sys system libraries
as well as the implementation of the network stacks which are located in
sys/net. Finally, the subdirectory
tests contains test applications,
including also a few expect scripts to automatically validate some of them.
Getting the source code
In order to clone the RIOT repository, you need the Git revision control system and run the following command:
git clone git://github.com/RIOT-OS/RIOT.git
The repository contains the kernel, support for different CPUs and platforms, device drivers, system libraries, and network stack implementations. In addition it comprises various example applications to demonstrate the usage of some important features.
It also provides you with useful tools like a terminal program and scripts to setup a toolchain.
Depending on the hardware you want to use, you need to first install a corresponding toolchain.
Platforms based on ARM
Platforms based on TI MSP430
Download and install GCC toolchain for MSP430 according to the information provided on the website.
For the native port
Create an application
Once you have set up the toolchain, you can create your own application. Apart from
the C file(s) containing your source code you need a Makefile. A template
Makefile is available in the
dist folder of the RIOT
Within your application's Makefile, you can define the target hardware as well as the modules you want to use.
Unless specified otherwise, make will create an elf-file as well as an Intel
hex file in the
bin folder of your application directory.
The build system
RIOT uses GNU make as build system. The simplest way to compile and link an application with RIOT, is to set up a Makefile providing at least the following variables:
and an instruction to include the
Makefile.include, located in RIOT's root
APPLICATION should contain the (unique) name of your application,
specifies the platform the application should be built for by default, and
RIOTBASE specifies the path to your copy of the RIOT repository (note, that
you may want to use
$(CURDIR) here, to give a relative path). You can use Make's
?= operator in order to allow overwriting variables from the command line. For
example, you can easily specify the target platform, using the sample Makefile,
by invoking make like this:
Besides typical targets like
doc, RIOT provides the special
term to invoke the configured flashing and terminal tools
for the specified platform. These targets use the variable
PORT for the serial
communication to the device. For the native port,
PORT has a special meaning,
it is used to identify the tap interface if the nativenet module is used.
debug can be used to invoke a debugger on some platforms.
For the native port the additional targets
all-valgrind target can be used instead of
all to build the application
for later use with the valgrind memory analyzer. This can be done with the
Some RIOT folders contain special Makefiles like
Makefile.dep. The first one can be included into other
Makefiles to define some standard targets. The files called
are used in
cpu to append target specific information to
INCLUDES, setting the include paths.
Makefile.dep serves to
By default a RIOT build comprises only the application code, the kernel, and platform specific code. In order to use additional modules, such as a particular device driver or a system library, you have to append the modules' names to the USEMODULE variable. For example, to build an application using the SHT11 temperature sensor and 6LoWPAN network stack, your Makefile needs to contain these lines:
USEMODULE += sht11 USEMODULE += sixlowpan
To contribute a new module to RIOT, your module's Makefile needs to set the
MODULE to a unique name. If the module depends on other modules, this
information needs to be added to RIOT's
The main function
After the board is initialized, RIOT starts two threads: the idle thread and the main thread. The idle thread has the lowest priority and will run, whenever no other thread is ready to run. It will automatically use the lowest possible power mode for the device. The main thread - configured with a default priority that is right in the middle between the lowest and the highest available priority - is the first thread that runs and calls the main function. This function needs to be defined by the application.
Choosing the right stack size
Choosing the right stack size for a new thread is not an easy, but a very crucial task. Since memory is usually a scarce resource in IoT scenarios, one must be careful not to assign too much stack to a single thread. However, if you allocate too little memory for a stack, your application will probably crash. The minimum stack size depends also on some RIOT internal structs and is hardware dependent. In order to help developers finding the right stack size, RIOT defines some typical stack sizes in
cpu-conf. (which should be provided by the implementation for all supported MCUs). The constants for these stack sizes are
and can be used by including
kernel.h. ARM based platforms additionally define
KERNEL_CONF_STACKSIZE_PRINTF_FLOAT, because newlibs printf implementation uses more memory for printing floating point numbers.
KERNEL_CONF_STACKSIZE_IDLE is the stack size for the idle thread and probably the smallest sensible stack size.
KERNEL_CONF_STACKSIZE_DEFAULT is a default size for any typical thread, not using printf.
KERNEL_CONF_STACKSIZE_PRINTF defines additional stack space needed if the thread needs to call printf (which requires additional memory when using newlib.
KERNEL_CONF_STACKSIZE_MAIN is the stack size for the main thread and probably a good size for your application. (Note, that on most non-newlib dependent platforms this will probably equal
Like any microkernel system, RIOT has an IPC API that enables data exchange between modules or a single module and the kernel. This API is documented in the doxygen documentation. The IPC can be used in several ways, such as synchronous or asynchronous, blocking or non-blocking, with or without a message queue. In the default case, a thread does not have a message queue. Hence, messages sent in a non-blocking manner are lost, when the target thread is not in receive mode. A thread may set up a message queue using the corresponding function, but has to provide the memory for this queue itself.
The transceiver module
The transceiver module is an abstraction layer and multiplexer between the
network stack and the radio driver. It runs in a single thread with the PID
transceiver_pid. It provides an IPC interface that enables to configure and
use available radio drivers, e.g. setting the radio channel or sending a packet.
A thread may also register at the transceiver module, in order to get notified
whenever a packet for a particular radio transceiver is received. The
notification message contains a pointer to the packet struct. After processing
the packet, the registered thread needs to decrease this struct's member
processing which acts as a semaphore for the packet's memory buffer.