Application Development¶
Note
In this document, we’ll assume your application directory is
<home>/app
, and that its build directory is
<home>/app/build
.
(These terms are defined in the following Overview.)
On Linux/macOS, <home> is equivalent to ~
, whereas on Windows it’s
%userprofile%
.
Overview¶
Zephyr’s build system is based on CMake.
The build system is application-centric, and requires Zephyr-based applications to initiate building the kernel source tree. The application build controls the configuration and build process of both the application and Zephyr itself, compiling them into a single binary.
Zephyr’s base directory hosts Zephyr’s own source code, its kernel configuration options, and its build definitions.
The files in the application directory link Zephyr with the application. This directory contains all application-specific files, such as configuration options and source code.
An application in its simplest form has the following contents:
<home>/app
├── CMakeLists.txt
├── prj.conf
└── src
└── main.c
These contents are:
CMakeLists.txt: This file tells the build system where to find the other application files, and links the application directory with Zephyr’s CMake build system. This link provides features supported by Zephyr’s build system, such as board-specific kernel configuration files, the ability to run and debug compiled binaries on real or emulated hardware, and more.
Kernel configuration files: An application typically provides a Kconfig configuration file (usually called
prj.conf
) that specifies application-specific values for one or more kernel configuration options. These application settings are merged with board-specific settings to produce a kernel configuration.See Kconfig Configuration below for more information.
Application source code files: An application typically provides one or more application-specific files, written in C or assembly language. These files are usually located in a sub-directory called
src
.
Once an application has been defined, you can use CMake to create project files for building it from a directory where you want to host these files. This is known as the build directory. Application build artifacts are always generated in a build directory; Zephyr does not support “in-tree” builds.
The following sections describe how to create, build, and run Zephyr applications, followed by more detailed reference material.
Source Tree Structure¶
Understanding the Zephyr source tree can be helpful in locating the code associated with a particular Zephyr feature.
At the top of the tree there are several files that are of importance:
CMakeLists.txt
The top-level file for the CMake build system, containing a lot of the logic required to build Zephyr.
Kconfig
The top-level Kconfig file, which refers to the file
Kconfig.zephyr
also found at the top-level directory.See the Kconfig section of the manual for detailed Kconfig documentation.
west.yml
The West (Zephyr’s meta-tool) manifest, listing the external repositories managed by the west command-line tool.
The Zephyr source tree also contains the following top-level directories, each of which may have one or more additional levels of subdirectories which are not described here.
arch
Architecture-specific kernel and system-on-chip (SoC) code. Each supported architecture (for example, x86 and ARM) has its own subdirectory, which contains additional subdirectories for the following areas:
architecture-specific kernel source files
architecture-specific kernel include files for private APIs
soc
SoC related code and configuration files.
boards
Board related code and configuration files.
doc
Zephyr technical documentation source files and tools used to generate the https://docs.zephyrproject.org web content.
drivers
Device driver code.
dts
devicetree source files used to describe non-discoverable board-specific hardware details.
include
Include files for all public APIs, except those defined under
lib
.kernel
Architecture-independent kernel code.
lib
Library code, including the minimal standard C library.
misc
Miscellaneous code that doesn’t belong to any of the other top-level directories.
samples
Sample applications that demonstrate the use of Zephyr features.
scripts
Various programs and other files used to build and test Zephyr applications.
cmake
Additional build scripts needed to build Zephyr.
subsys
Subsystems of Zephyr, including:
USB device stack code.
Networking code, including the Bluetooth stack and networking stacks.
File system code.
Bluetooth host and controller
tests
Test code and benchmarks for Zephyr features.
share
Additional architecture independent data. Currently containing Zephyr CMake package.
Example standalone application¶
A reference standalone application contained in its own Git repository can be found in the Example Application repository. It can be used as a reference on how to structure out-of-tree, Zephyr-based applications using the T2 star topology. It also demonstrates the out-of-tree use of features commonly used in applications such as:
Custom boards
Custom devicetree bindings
Custom drivers
Continuous Integration (CI) setup
Creating an Application¶
Follow these steps to create a new application directory. (Refer to the Example Application repository for a reference standalone application in its own Git repository or to Samples and Demos for existing applications provided as part of Zephyr.)
Create an application directory on your workstation computer, outside of the Zephyr base directory. Usually you’ll want to create it somewhere under your user’s home directory.
For example, in a Unix shell or Windows
cmd.exe
prompt, navigate to where you want to create your application, then enter:mkdir app
Warning
Building Zephyr or creating an application in a directory with spaces anywhere on the path is not supported. So the Windows path
C:\Users\YourName\app
will work, butC:\Users\Your Name\app
will not.It’s recommended to place all application source code in a subdirectory named
src
. This makes it easier to distinguish between project files and sources.Continuing the previous example, enter:
cd app mkdir src
Place your application source code in the
src
sub-directory. For this example, we’ll assume you created a file namedsrc/main.c
.Create a file named
CMakeLists.txt
in theapp
directory with the following contents:# Find Zephyr. This also loads Zephyr's build system. cmake_minimum_required(VERSION 3.13.1) find_package(Zephyr) project(my_zephyr_app) # Add your source file to the "app" target. This must come after # find_package(Zephyr) which defines the target. target_sources(app PRIVATE src/main.c)
find_package(Zephyr)
sets the minimum CMake version and pulls in the Zephyr build system, which creates a CMake target namedapp
(see Zephyr CMake Package). Adding sources to this target is how you include them in the build.Note
cmake_minimum_required()
is also invoked by the Zephyr package. The most recent of the two versions will be enforced by CMake.Set Kconfig configuration options. See Kconfig Configuration.
Configure any devicetree overlays needed by your application. See Set devicetree overlays.
Note
include($ENV{ZEPHYR_BASE}/cmake/app/boilerplate.cmake NO_POLICY_SCOPE)
is still supported for backward compatibility with older applications.
Including boilerplate.cmake
directly in the sample still requires to run
source zephyr-env.sh
or execute zephyr-env.cmd
before building the
application.
Setting Variables¶
Option 1: Just Once¶
To set the environment variable MY_VARIABLE
to foo
for the
lifetime of your current terminal window:
# Linux and macOS
export MY_VARIABLE=foo
# Windows
set MY_VARIABLE=foo
Warning
This is best for experimentation. If you close your terminal window, use another terminal window or tab, restart your computer, etc., this setting will be lost forever.
Using options 2 or 3 is recommended if you want to keep using the setting.
Option 2: In all Terminals¶
macOS and Linux:
Add the export MY_VARIABLE=foo
line to your shell’s startup script in your
home directory. For Bash, this is usually ~/.bashrc
on Linux or
~/.bash_profile
on macOS. Changes in these startup scripts don’t
affect shell instances already started; try opening a new terminal window to get
the new settings.
Windows:
You can use the setx
program in cmd.exe
or the third-party
RapidEE program.
To use setx
, type this command, then close the terminal window. Any new
cmd.exe
windows will have MY_VARIABLE
set to foo
.
setx MY_VARIABLE foo
To install RapidEE, a freeware graphical environment variable editor, using Chocolatey in an Administrator command prompt:
choco install rapidee
You can then run rapidee
from your terminal to launch the program and set
environment variables. Make sure to use the “User” environment variables area
– otherwise, you have to run RapidEE as administrator. Also make sure to save
your changes by clicking the Save button at top left before exiting.Settings
you make in RapidEE will be available whenever you open a new terminal window.
Option 3: Using zephyrrc
files¶
Choose this option if you don’t want to make the variable’s setting available to all of your terminals, but still want to save the value for loading into your environment when you are using Zephyr.
macOS and Linux:
Create a file named ~/.zephyrrc
if it doesn’t exist, then add this line
to it:
export MY_VARIABLE=foo
To get this value back into your current terminal environment, you must run
source zephyr-env.sh
from the main zephyr
repository. Among other
things, this script sources ~/.zephyrrc
.
The value will be lost if you close the window, etc.; run source
zephyr-env.sh
again to get it back.
Windows:
Add the line set MY_VARIABLE=foo
to the file
%userprofile%\zephyrrc.cmd
using a text editor such as Notepad to save
the value.
To get this value back into your current terminal environment, you must run
zephyr-env.cmd
in a cmd.exe
window after changing directory to the main
zephyr
repository. Among other things, this script runs
%userprofile%\zephyrrc.cmd
.
The value will be lost if you close the window, etc.; run zephyr-env.cmd
again to get it back.
These scripts:
set
ZEPHYR_BASE
(see below) to the location of the zephyr repositoryadds some Zephyr-specific locations (such as zephyr’s
scripts
directory) to yourPATH
environment variableloads any settings from the
zephyrrc
files described above in Option 3: Using zephyrrc files.
You can thus use them any time you need any of these settings.
Option 4: Using Zephyr Build Configuration CMake package¶
Choose this option if you want to make those variable settings shared among all users of your project.
Using a Zephyr Build Configuration CMake package allows you to commit the shared settings into the repository, so that all users can share them.
It also removes the need for running source zephyr-env.sh
or
zephyr-env.cmd
when opening a new terminal.
Important Build System Variables¶
You can control the Zephyr build system using many variables. This section describes the most important ones that every Zephyr developer should know about.
Note
The variables BOARD, CONF_FILE, and DTC_OVERLAY_FILE can be supplied to the build system in 3 ways (in order of precedence):
As a parameter to the
west build
orcmake
invocation via the-D
command-line switch. If you have multiple overlay files, you should use quotations,"file1.overlay;file2.overlay"
As a
set(<VARIABLE> <VALUE>)
statement in yourCMakeLists.txt
ZEPHYR_BASE: Zephyr base variable used by the build system.
find_package(Zephyr)
will automatically set this as a cached CMake variable. ButZEPHYR_BASE
can also be set as an environment variable in order to force CMake to use a specific Zephyr installation.BOARD: Selects the board that the application’s build will use for the default configuration. See Supported Boards for built-in boards, and Board Porting Guide for information on adding board support.
CONF_FILE: Indicates the name of one or more Kconfig configuration fragment files. Multiple filenames can be separated with either spaces or semicolons. Each file includes Kconfig configuration values that override the default configuration values.
See The Initial Configuration for more information.
OVERLAY_CONFIG: Additional Kconfig configuration fragment files. Multiple filenames can be separated with either spaces or semicolons. This can be useful in order to leave CONF_FILE at its default value, but “mix in” some additional configuration options.
DTC_OVERLAY_FILE: One or more devicetree overlay files to use. Multiple files can be separated with semicolons. See Set devicetree overlays for examples and Introduction to devicetree for information about devicetree and Zephyr.
ZEPHYR_MODULES: A CMake list containing absolute paths of additional directories with source code, Kconfig, etc. that should be used in the application build. See Modules (External projects) for details.
Application CMakeLists.txt¶
Every application must have a CMakeLists.txt
file. This file is the
entry point, or top level, of the build system. The final zephyr.elf
image contains both the application and the kernel libraries.
This section describes some of what you can do in your CMakeLists.txt
.
Make sure to follow these steps in order.
If you only want to build for one board, add the name of the board configuration for your application on a new line. For example:
set(BOARD qemu_x86)
Refer to Supported Boards for more information on available boards.
The Zephyr build system determines a value for BOARD by checking the following, in order (when a BOARD value is found, CMake stops looking further down the list):
Any previously used value as determined by the CMake cache takes highest precedence. This ensures you don’t try to run a build with a different BOARD value than you set during the build configuration step.
Any value given on the CMake command line (directly or indirectly via
west build
) using-DBOARD=YOUR_BOARD
will be checked for and used next.If an environment variable
BOARD
is set, its value will then be used.Finally, if you set
BOARD
in your applicationCMakeLists.txt
as described in this step, this value will be used.
If your application uses a configuration file or files other than the usual
prj.conf
(orprj_YOUR_BOARD.conf
, whereYOUR_BOARD
is a board name), add lines setting the CONF_FILE variable to these files appropriately. If multiple filenames are given, separate them by a single space or semicolon. CMake lists can be used to build up configuration fragment files in a modular way when you want to avoid setting CONF_FILE in a single place. For example:set(CONF_FILE "fragment_file1.conf") list(APPEND CONF_FILE "fragment_file2.conf")
See The Initial Configuration for more information.
If your application uses devicetree overlays, you may need to set DTC_OVERLAY_FILE. See Set devicetree overlays.
If your application has its own kernel configuration options, create a
Kconfig
file in the same directory as your application’sCMakeLists.txt
.See the Kconfig section of the manual for detailed Kconfig documentation.
An (unlikely) advanced use case would be if your application has its own unique configuration options that are set differently depending on the build configuration.
If you just want to set application specific values for existing Zephyr configuration options, refer to the CONF_FILE description above.
Structure your
Kconfig
file like this:# SPDX-License-Identifier: Apache-2.0 mainmenu "Your Application Name" # Your application configuration options go here # Sources Kconfig.zephyr in the Zephyr root directory. # # Note: All 'source' statements work relative to the Zephyr root directory (due # to the $srctree environment variable being set to $ZEPHYR_BASE). If you want # to 'source' relative to the current Kconfig file instead, use 'rsource' (or a # path relative to the Zephyr root). source "Kconfig.zephyr"
Note
Environment variables in
source
statements are expanded directly, so you do not need to define anoption env="ZEPHYR_BASE"
Kconfig “bounce” symbol. If you use such a symbol, it must have the same name as the environment variable.See Kconfig extensions for more information.
The
Kconfig
file is automatically detected when placed in the application directory, but it is also possible for it to be found elsewhere if the CMake variable KCONFIG_ROOT is set with an absolute path.Specify that the application requires Zephyr on a new line, after any lines added from the steps above:
find_package(Zephyr) project(my_zephyr_app)
Note
find_package(Zephyr REQUIRED HINTS $ENV{ZEPHYR_BASE})
can be used if enforcing a specific Zephyr installation by explicitly setting theZEPHYR_BASE
environment variable should be supported. All samples in Zephyr supports theZEPHYR_BASE
environment variable.Now add any application source files to the ‘app’ target library, each on their own line, like so:
target_sources(app PRIVATE src/main.c)
Below is a simple example CMakeList.txt
:
set(BOARD qemu_x86)
find_package(Zephyr)
project(my_zephyr_app)
target_sources(app PRIVATE src/main.c)
The Cmake property HEX_FILES_TO_MERGE
leverages the application configuration provided by
Kconfig and CMake to let you merge externally built hex files
with the hex file generated when building the Zephyr application.
For example:
set_property(GLOBAL APPEND PROPERTY HEX_FILES_TO_MERGE
${app_bootloader_hex}
${PROJECT_BINARY_DIR}/${KERNEL_HEX_NAME}
${app_provision_hex})
CMakeCache.txt¶
CMake uses a CMakeCache.txt file as persistent key/value string storage used to cache values between runs, including compile and build options and paths to library dependencies. This cache file is created when CMake is run in an empty build folder.
For more details about the CMakeCache.txt file see the official CMake documentation runningcmake .
Application Configuration¶
Kconfig Configuration¶
Application configuration options are usually set in prj.conf
in the
application directory. For example, C++ support could be enabled with this
assignment:
CONFIG_CPLUSPLUS=y
Looking at existing samples is a good way to get started.
See Setting Kconfig configuration values for detailed documentation on setting Kconfig configuration values. The The Initial Configuration section on the same page explains how the initial configuration is derived. See configuration_options for a complete list of configuration options. See Hardening Tool for security information related with Kconfig options.
The other pages in the Kconfig section of the manual are also worth going through, especially if you planning to add new configuration options.
Devicetree Overlays¶
Application-Specific Code¶
Application-specific source code files are normally added to the
application’s src
directory. If the application adds a large
number of files the developer can group them into sub-directories
under src
, to whatever depth is needed.
Application-specific source code should not use symbol name prefixes that have been reserved by the kernel for its own use. For more information, see Naming Conventions.
Third-party Library Code¶
It is possible to build library code outside the application’s src
directory but it is important that both application and library code targets
the same Application Binary Interface (ABI). On most architectures there are
compiler flags that control the ABI targeted, making it important that both
libraries and applications have certain compiler flags in common. It may also
be useful for glue code to have access to Zephyr kernel header files.
To make it easier to integrate third-party components, the Zephyr
build system has defined CMake functions that give application build
scripts access to the zephyr compiler options. The functions are
documented and defined in cmake/extensions.cmake
and follow the naming convention zephyr_get_<type>_<format>
.
The following variables will often need to be exported to the third-party build system.
CMAKE_C_COMPILER
,CMAKE_AR
.ARCH
andBOARD
, together with several variables that identify the Zephyr kernel version.
samples/application_development/external_lib is a sample project that demonstrates some of these features.
Building an Application¶
The Zephyr build system compiles and links all components of an application into a single application image that can be run on simulated hardware or real hardware.
Like any other CMake-based system, the build process takes place in
two stages. First, build files (also known as a buildsystem)
are generated using the cmake
command-line tool while specifying a
generator. This generator determines the native build tool the buildsystem
will use in the second stage.
The second stage runs the native build tool to actually build the
source files and generate an image. To learn more about these concepts refer to
the CMake introduction in the official CMake documentation.
Although the default build tool in Zephyr is west, Zephyr’s
meta-tool, which invokes cmake
and the underlying build tool (ninja
or
make
) behind the scenes, you can also choose to invoke cmake
directly if
you prefer. On Linux and macOS you can choose between the make
and
ninja
generators (i.e. build tools), whereas on Windows you need to use ninja
,
since make
is not supported on this platform.
For simplicity we will use ninja
throughout this guide, and if you
choose to use west build
to build your application know that it will
default to ninja
under the hood.
As an example, let’s build the Hello World sample for the reel_board
:
Using west:
west build -b reel_board samples/hello_world
Using CMake and ninja:
# Use cmake to configure a Ninja-based buildsystem:
cmake -B build -GNinja -DBOARD=reel_board samples/hello_world
# Now run ninja on the generated build system:
ninja -C build
On Linux and macOS, you can also build with make
instead of ninja
:
Using west:
to use
make
just once, add-- -G"Unix Makefiles"
to the west build command line; see the west build documentation for an example.to use
make
by default from now on, runwest config build.generator "Unix Makefiles"
.
Using CMake directly:
# Use cmake to configure a Make-based buildsystem:
cmake -B build -DBOARD=reel_board samples/hello_world
# Now run ninja on the generated build system:
make -C build
Basics¶
Note
In the below example, west
is used outside of a west workspace. For this
to work, you must set the ZEPHYR_BASE
environment variable to the path
of your zephyr git repository, using one of the methods on the
Environment Variables page.
Navigate to the application directory
<home>/app
.Enter the following commands to build the application’s
zephyr.elf
image for the board specified in the command-line parameters:Using west:
west build -b <board>
Using CMake and ninja:
mkdir build && cd build # Use cmake to configure a Ninja-based buildsystem: cmake -GNinja -DBOARD=<board> .. # Now run ninja on the generated build system: ninja
If desired, you can build the application using the configuration settings specified in an alternate
.conf
file using theCONF_FILE
parameter. These settings will override the settings in the application’s.config
file or its default.conf
file. For example:Using west:
west build -b <board> -- -DCONF_FILE=prj.alternate.conf
Using CMake and ninja:
mkdir build && cd build cmake -GNinja -DBOARD=<board> -DCONF_FILE=prj.alternate.conf .. ninja
As described in the previous section, you can instead choose to permanently set the board and configuration settings by either exporting BOARD and CONF_FILE environment variables or by setting their values in your
CMakeLists.txt
usingset()
statements. Additionally,west
allows you to set a default board.
Build Directory Contents¶
When using the Ninja generator a build directory looks like this:
<home>/app/build
├── build.ninja
├── CMakeCache.txt
├── CMakeFiles
├── cmake_install.cmake
├── rules.ninja
└── zephyr
The most notable files in the build directory are:
build.ninja
, which can be invoked to build the application.A
zephyr
directory, which is the working directory of the generated build system, and where most generated files are created and stored.
After running ninja
, the following build output files will be written to
the zephyr
sub-directory of the build directory. (This is not the
Zephyr base directory, which contains the Zephyr source code etc. and is
described above.)
.config
, which contains the configuration settings used to build the application.Note
The previous version of
.config
is saved to.config.old
whenever the configuration is updated. This is for convenience, as comparing the old and new versions can be handy.Various object files (
.o
files and.a
files) containing compiled kernel and application code.zephyr.elf
, which contains the final combined application and kernel binary. Other binary output formats, such as.hex
and.bin
, are also supported.
Rebuilding an Application¶
Application development is usually fastest when changes are continually tested. Frequently rebuilding your application makes debugging less painful as the application becomes more complex. It’s usually a good idea to rebuild and test after any major changes to the application’s source files, CMakeLists.txt files, or configuration settings.
Important
The Zephyr build system rebuilds only the parts of the application image potentially affected by the changes. Consequently, rebuilding an application is often significantly faster than building it the first time.
Sometimes the build system doesn’t rebuild the application correctly because it fails to recompile one or more necessary files. You can force the build system to rebuild the entire application from scratch with the following procedure:
Open a terminal console on your host computer, and navigate to the build directory
<home>/app/build
.Enter one of the following commands, depending on whether you want to use
west
orcmake
directly to delete the application’s generated files, except for the.config
file that contains the application’s current configuration information.west build -t clean
or
ninja clean
Alternatively, enter one of the following commands to delete all generated files, including the
.config
files that contain the application’s current configuration information for those board types.west build -t pristine
or
ninja pristine
If you use west, you can take advantage of its capability to automatically make the build folder pristine whenever it is required.
Rebuild the application normally following the steps specified in Building an Application above.
Building for a board revision¶
The Zephyr build system has support for specifying multiple hardware revisions of a single board with small variations. Using revisions allows the board support files to make minor adjustments to a board configuration without duplicating all the files described in Create your board directory for each revision.
To build for a particular revision, use <board>@<revision>
instead of plain
<board>
. For example:
Using west:
west build -b <board>@<revision>
Using CMake and ninja:
mkdir build && cd build
cmake -GNinja -DBOARD=<board>@<revision> ..
ninja
Check your board’s documentation for details on whether it has multiple revisions, and what revisions are supported.
When targeting a board revision, the active revision will be printed at CMake configure time, like this:
-- Board: plank, Revision: 1.5.0
Run an Application¶
An application image can be run on a real board or emulated hardware.
Running on a Board¶
Most boards supported by Zephyr let you flash a compiled binary using
the flash
target to copy the binary to the board and run it.
Follow these instructions to flash and run an application on real
hardware:
Build your application, as described in Building an Application.
Make sure your board is attached to your host computer. Usually, you’ll do this via USB.
Run one of these console commands from the build directory,
<home>/app/build
, to flash the compiled Zephyr image and run it on your board:west flash
or
ninja flash
The Zephyr build system integrates with the board support files to use hardware-specific tools to flash the Zephyr binary to your hardware, then run it.
Each time you run the flash command, your application is rebuilt and flashed again.
In cases where board support is incomplete, flashing via the Zephyr build system may not be supported. If you receive an error message about flash support being unavailable, consult your board’s documentation for additional information on how to flash your board.
Note
When developing on Linux, it’s common to need to install board-specific udev rules to enable USB device access to your board as a non-root user. If flashing fails, consult your board’s documentation to see if this is necessary.
Running in an Emulator¶
The kernel has built-in emulator support for QEMU (on Linux/macOS only, this is not yet supported on Windows). It allows you to run and test an application virtually, before (or in lieu of) loading and running it on actual target hardware. Follow these instructions to run an application via QEMU:
Build your application for one of the QEMU boards, as described in Building an Application.
For example, you could set
BOARD
to:qemu_x86
to emulate running on an x86-based boardqemu_cortex_m3
to emulate running on an ARM Cortex M3-based board
Run one of these console commands from the build directory,
<home>/app/build
, to run the Zephyr binary in QEMU:west build -t run
or
ninja run
Press Ctrl A, X to stop the application from running in QEMU.
The application stops running and the terminal console prompt redisplays.
Each time you execute the run command, your application is rebuilt and run again.
Note
If the (Linux only) Zephyr SDK is installed, the run
target will use the SDK’s QEMU binary by default. To use another version of
QEMU, set the environment variable QEMU_BIN_PATH
to the path of the QEMU binary you want to use instead.
Application Debugging¶
This section is a quick hands-on reference to start debugging your application with QEMU. Most content in this section is already covered in QEMU and GNU_Debugger reference manuals.
In this quick reference, you’ll find shortcuts, specific environmental variables, and parameters that can help you to quickly set up your debugging environment.
The simplest way to debug an application running in QEMU is using the GNU Debugger and setting a local GDB server in your development system through QEMU.
You will need an Executable and Linkable Format (ELF) binary image for
debugging purposes. The build system generates the image in the build
directory. By default, the kernel binary name is
zephyr.elf
. The name can be changed using a Kconfig option.
We will use the standard 1234 TCP port to open a GDB server instance. This port number can be changed for a port that best suits the development environment.
You can run QEMU to listen for a “gdb connection” before it starts executing any code to debug it.
qemu -s -S <image>
will setup Qemu to listen on port 1234 and wait for a GDB connection to it.
The options used above have the following meaning:
-S
Do not start CPU at startup; rather, you must type ‘c’ in the monitor.-s
Shorthand for-gdb tcp::1234
: open a GDB server on TCP port 1234.
To debug with QEMU and to start a GDB server and wait for a remote connect, run either of the following inside the build directory of an application:
ninja debugserver
The build system will start a QEMU instance with the CPU halted at startup and with a GDB server instance listening at the TCP port 1234.
Using a local GDB configuration .gdbinit
can help initialize your GDB
instance on every run.
In this example, the initialization file points to the GDB server instance.
It configures a connection to a remote target at the local host on the TCP
port 1234. The initialization sets the kernel’s root directory as a
reference.
The .gdbinit
file contains the following lines:
target remote localhost:1234
dir ZEPHYR_BASE
Note
Substitute the correct ZEPHYR_BASE for your system.
Execute the application to debug from the same directory that you chose for
the gdbinit
file. The command can include the --tui
option
to enable the use of a terminal user interface. The following commands
connects to the GDB server using gdb
. The command loads the symbol
table from the elf binary file. In this example, the elf binary file name
corresponds to zephyr.elf
file:
..../path/to/gdb --tui zephyr.elf
Note
The GDB version on the development system might not support the –tui option. Please make sure you use the GDB binary from the SDK which corresponds to the toolchain that has been used to build the binary.
If you are not using a .gdbinit file, issue the following command inside GDB to connect to the remote GDB server on port 1234:
(gdb) target remote localhost:1234
Finally, the command below connects to the GDB server using the Data
Displayer Debugger (ddd
). The command loads the symbol table from the
elf binary file, in this instance, the zephyr.elf
file.
The DDD may not be installed in your
development system by default. Follow your system instructions to install
it. For example, use sudo apt-get install ddd
on an Ubuntu system.
ddd --gdb --debugger "gdb zephyr.elf"
Both commands execute the gdb. The command name might change depending on the toolchain you are using and your cross-development tools.
Custom Board, Devicetree and SOC Definitions¶
In cases where the board or platform you are developing for is not yet supported by Zephyr, you can add board, Devicetree and SOC definitions to your application without having to add them to the Zephyr tree.
The structure needed to support out-of-tree board and SOC development is similar to how boards and SOCs are maintained in the Zephyr tree. By using this structure, it will be much easier to upstream your platform related work into the Zephyr tree after your initial development is done.
Add the custom board to your application or a dedicated repository using the following structure:
boards/
soc/
CMakeLists.txt
prj.conf
README.rst
src/
where the boards
directory hosts the board you are building for:
.
├── boards
│ └── x86
│ └── my_custom_board
│ ├── doc
│ │ └── img
│ └── support
└── src
and the soc
directory hosts any SOC code. You can also have boards that are
supported by a SOC that is available in the Zephyr tree.
Boards¶
Use the proper architecture folder name (e.g., x86
, arm
, etc.)
under boards
for my_custom_board
. (See Supported Boards for a
list of board architectures.)
Documentation (under doc/
) and support files (under support/
) are optional, but
will be needed when submitting to Zephyr.
The contents of my_custom_board
should follow the same guidelines for any
Zephyr board, and provide the following files:
my_custom_board_defconfig
my_custom_board.dts
my_custom_board.yaml
board.cmake
board.h
CMakeLists.txt
doc/
dts_fixup.h
Kconfig.board
Kconfig.defconfig
pinmux.c
support/
Once the board structure is in place, you can build your application
targeting this board by specifying the location of your custom board
information with the -DBOARD_ROOT
parameter to the CMake
build system:
Using west:
west build -b <board name> -- -DBOARD_ROOT=<path to boards>
Using CMake and ninja:
cmake -B build -GNinja -DBOARD=<board name> -DBOARD_ROOT=<path to boards> .
ninja -C build
This will use your custom board configuration and will generate the Zephyr binary into your application directory.
You can also define the BOARD_ROOT
variable in the application
CMakeLists.txt
file. Make sure to do so before pulling in the Zephyr
boilerplate with find_package(Zephyr ...)
.
Note
When specifying BOARD_ROOT
in a CMakeLists.txt, then an absolute path must
be provided, for example list(APPEND BOARD_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/<extra-board-root>
.
When using -DBOARD_ROOT=<board-root>
both absolute and relative paths can
be used. Relative paths are treated relatively to the application directory.
SOC Definitions¶
Similar to board support, the structure is similar to how SOCs are maintained in the Zephyr tree, for example:
soc
└── arm
└── st_stm32
├── common
└── stm32l0
The file soc/Kconfig will create the top-level
SoC/CPU/Configuration Selection
menu in Kconfig.
Out of tree SoC definitions can be added to this menu using the SOC_ROOT
CMake variable. This variable contains a semicolon-separated list of directories
which contain SoC support files.
Following the structure above, the following files can be added to load more SoCs into the menu.
soc
└── arm
└── st_stm32
├── Kconfig
├── Kconfig.soc
└── Kconfig.defconfig
The Kconfig files above may describe the SoC or load additional SoC Kconfig files.
An example of loading stm31l0
specific Kconfig files in this structure:
soc
└── arm
└── st_stm32
├── Kconfig.soc
└── stm32l0
└── Kconfig.series
can be done with the following content in st_stm32/Kconfig.soc
:
rsource "*/Kconfig.series"
Once the SOC structure is in place, you can build your application
targeting this platform by specifying the location of your custom platform
information with the -DSOC_ROOT
parameter to the CMake
build system:
Using west:
west build -b <board name> -- -DSOC_ROOT=<path to soc> -DBOARD_ROOT=<path to boards>
Using CMake and ninja:
cmake -B build -GNinja -DBOARD=<board name> -DSOC_ROOT=<path to soc> -DBOARD_ROOT=<path to boards> .
ninja -C build
This will use your custom platform configurations and will generate the Zephyr binary into your application directory.
See Build settings for information on setting SOC_ROOT in a module’s
zephyr/module.yml
file.
Or you can define the SOC_ROOT
variable in the application
CMakeLists.txt
file. Make sure to do so before pulling in the
Zephyr boilerplate with find_package(Zephyr ...)
.
Note
When specifying SOC_ROOT
in a CMakeLists.txt, then an absolute path must
be provided, for example list(APPEND SOC_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/<extra-soc-root>
.
When using -DSOC_ROOT=<soc-root>
both absolute and relative paths can be
used. Relative paths are treated relatively to the application directory.
Devicetree Definitions¶
Devicetree directory trees are found in APPLICATION_SOURCE_DIR
,
BOARD_DIR
, and ZEPHYR_BASE
, but additional trees, or DTS_ROOTs,
can be added by creating this directory tree:
include/
dts/common/
dts/arm/
dts/
dts/bindings/
Where ‘arm’ is changed to the appropriate architecture. Each directory is optional. The binding directory contains bindings and the other directories contain files that can be included from DT sources.
Once the directory structure is in place, you can use it by specifying
its location through the DTS_ROOT
CMake Cache variable:
Using west:
west build -b <board name> -- -DDTS_ROOT=<path to dts root>
Using CMake and ninja:
cmake -B build -GNinja -DBOARD=<board name> -DDTS_ROOT=<path to dts root> .
ninja -C build
You can also define the variable in the application CMakeLists.txt
file. Make sure to do so before pulling in the Zephyr boilerplate with
find_package(Zephyr ...)
.
Note
When specifying DTS_ROOT
in a CMakeLists.txt, then an absolute path must
be provided, for example list(APPEND DTS_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/<extra-dts-root>
.
When using -DDTS_ROOT=<dts-root>
both absolute and relative paths can be
used. Relative paths are treated relatively to the application directory.
Devicetree source are passed through the C preprocessor, so you can
include files that can be located in a DTS_ROOT
directory. By
convention devicetree include files have a .dtsi
extension.
You can also use the preprocessor to control the content of a devicetree
file, by specifying directives through the DTS_EXTRA_CPPFLAGS
CMake
Cache variable:
Using west:
west build -b <board name> -- -DDTS_EXTRA_CPPFLAGS=-DTEST_ENABLE_FEATURE
Using CMake and ninja:
cmake -B build -GNinja -DBOARD=<board name> -DDTS_EXTRA_CPPFLAGS=-DTEST_ENABLE_FEATURE .
ninja -C build
Debug with Eclipse¶
Overview¶
CMake supports generating a project description file that can be imported into the Eclipse Integrated Development Environment (IDE) and used for graphical debugging.
The GNU MCU Eclipse plug-ins provide a mechanism to debug ARM projects in Eclipse with pyOCD, Segger J-Link, and OpenOCD debugging tools.
The following tutorial demonstrates how to debug a Zephyr application in Eclipse with pyOCD in Windows. It assumes you have already installed the GCC ARM Embedded toolchain and pyOCD.
Set Up the Eclipse Development Environment¶
Download and install Eclipse IDE for C/C++ Developers.
In Eclipse, install the GNU MCU Eclipse plug-ins by opening the menu
Window->Eclipse Marketplace...
, searching forGNU MCU Eclipse
, and clickingInstall
on the matching result.Configure the path to the pyOCD GDB server by opening the menu
Window->Preferences
, navigating toMCU
, and setting theGlobal pyOCD Path
.
Generate and Import an Eclipse Project¶
Set up a GNU Arm Embedded toolchain as described in 3rd Party Toolchains.
Navigate to a folder outside of the Zephyr tree to build your application.
# On Windows cd %userprofile%
Note
If the build directory is a subdirectory of the source directory, as is usually done in Zephyr, CMake will warn:
“The build directory is a subdirectory of the source directory.
This is not supported well by Eclipse. It is strongly recommended to use a build directory which is a sibling of the source directory.”
Configure your application with CMake and build it with ninja. Note the different CMake generator specified by the
-G"Eclipse CDT4 - Ninja"
argument. This will generate an Eclipse project description file,.project
, in addition to the usual ninja build files.Using west:
west build -b frdm_k64f %ZEPHYR_BASE%\samples\synchronization -- -G"Eclipse CDT4 - Ninja"
Using CMake and ninja:
cmake -B build -GNinja -DBOARD=frdm_k64f -G"Eclipse CDT4 - Ninja" %ZEPHYR_BASE%\samples\synchronization ninja -C build
In Eclipse, import your generated project by opening the menu
File->Import...
and selecting the optionExisting Projects into Workspace
. Browse to your application build directory in the choice,Select root directory:
. Check the box for your project in the list of projects found and click theFinish
button.
Create a Debugger Configuration¶
Open the menu
Run->Debug Configurations...
.Select
GDB PyOCD Debugging
, click theNew
button, and configure the following options:In the Main tab:
Project: my_zephyr_app@build
C/C++ Application:
zephyr/zephyr.elf
In the Debugger tab:
pyOCD Setup
Executable path:
$pyocd_path\$pyocd_executable
Uncheck “Allocate console for semihosting”
Board Setup
Bus speed: 8000000 Hz
Uncheck “Enable semihosting”
GDB Client Setup
Executable path example (use your
GNUARMEMB_TOOLCHAIN_PATH
):C:\gcc-arm-none-eabi-6_2017-q2-update\bin\arm-none-eabi-gdb.exe
In the SVD Path tab:
File path:
<workspace top>\modules\hal\nxp\mcux\devices\MK64F12\MK64F12.xml
Note
This is optional. It provides the SoC’s memory-mapped register addresses and bitfields to the debugger.
Click the
Debug
button to start debugging.
RTOS Awareness¶
Support for Zephyr RTOS awareness is implemented in pyOCD v0.11.0 and later. It is compatible with GDB PyOCD Debugging in Eclipse, but you must enable CONFIG_DEBUG_THREAD_INFO=y in your application.