Modules (External projects)¶
Zephyr relies on the source code of several externally maintained projects in order to avoid reinventing the wheel and to reuse as much well-established, mature code as possible when it makes sense. In the context of Zephyr’s build system those are called modules. These modules must be integrated with the Zephyr build system, as described in more detail in other sections on this page.
To be classified as a candidate for being included in the default list of modules, an external project is required to have its own life-cycle outside the Zephyr Project, that is, reside in its own repository, and have its own contribution and maintenance workfow and release process. Zephyr modules should not contain code that is written exclusively for Zephyr. Instead, such code should be contributed to the main zephyr tree.
Modules to be included in the default manifest of the Zephyr project need to provide functionality or features endorsed and approved by the project Technical Steering Committee and should comply with the module licensing requirements and contribute guidelines. They should also have a Zephyr developer that is committed to maintain the module codebase.
Zephyr depends on several categories of modules, including but not limited to:
Debugger integration
Silicon vendor Hardware Abstraction Layers (HALs)
Cryptography libraries
File Systems
Inter-Process Communication (IPC) libraries
This page summarizes a list of policies and best practices which aim at better organizing the workflow in Zephyr modules.
Module Repositories¶
All modules included in the default manifest shall be hosted in repositories under the zephyrproject-rtos GitHub organization.
The module repository codebase shall include a module.yml file in a
zephyr/
folder at the root of the repository.Module repository names should follow the convention of using lowercase letters and dashes instead of underscores. This rule will apply to all new module repositories, except for repositories that are directly tracking external projects (hosted in Git repositories); such modules may be named as their external project counterparts.
Note
Existing module repositories that do not conform to the above convention do not need to be renamed to comply with the above convention.
Modules should use “zephyr” as the default name for the repository main branch. Branches for specific purposes, for example, a module branch for an LTS Zephyr version, shall have names starting with the ‘zephyr_’ prefix.
If the module has an external (upstream) project repository, the module repository should preserve the upstream repository folder structure.
Note
It is not required in module repositories to maintain a ‘master’ branch mirroring the master branch of the external repository. It is not recommended as this may generate confusion around the module’s main branch, which should be ‘zephyr’.
Synchronizing with upstream¶
It is preferred to synchronize a module respository with the latest stable release of the corresponding external project. It is permitted, however, to update a Zephyr module repository with the latest development branch tip, if this is required to get important updates in the module codebase. When synchronizing a module with upstream it is mandatory to document the rationale for performing the particular update.
Requirements for allowed practices¶
Changes to the main branch of a module repository, including synchronization with upstream code base, may only be applied via pull requests. These pull requests shall be verifiable by Zephyr CI and mergeable (e.g. with the Rebase and merge, or Create a merge commit option using Github UI). This ensures that the incoming changes are always reviewable, and the downstream module repository history is incremental (that is, existing commits, tags, etc. are always preserved). This policy also allows to run Zephyr CI, git lint, identity, and license checks directly on the set of changes that are to be brought into the module repository.
Note
Force-pushing to a module’s main branch is not allowed.
Allowed practices¶
The following practices conform to the above requirements and should be followed in all modules repositories. It is up to the module code owner to select the preferred synchronization practice, however, it is required that the selected practice is consistently followed in the respective module repository.
Updating modules with a diff from upstream: Upstream changes brought as a single snapshot commit (manual diff) in a pull request against the module’s main branch, which may be merged using the Rebase & merge operation. This approach is simple and should be applicable to all modules with the downside of supressing the upstream history in the module repository.
Note
The above practice is the only allowed practice in modules where the external project is not hosted in an upstream Git repository.
The commit message is expected to identify the upstream project URL, the version to which the module is updated (upstream version, tag, commit SHA, if applicable, etc.), and the reason for the doing the update.
Updating modules by merging the upstream branch: Upstream changes brought in by performing a Git merge of the intended upstream branch (e.g. main branch, latest release branch, etc.) submitting the result in pull request against the module main branch, and merging the pull request using the Create a merge commit operation. This approach is applicable to modules with an upstream project Git repository. The main advantages of this approach is that the upstream repository history (that is, the original commit SHAs) is preserved in the module repository. The downside of this approach is that two additional merge commits are generated in the downstream main branch.
Contributing to Zephyr modules¶
Individual Roles & Responsibilities¶
To facilitate management of Zephyr module repositories, the following individual roles are defined.
Administrator: Each Zephyr module shall have an administrator who is responsible for managing access to the module repository, for example, for adding individuals as Collaborators in the repository at the request of the module owner. Module administrators are members of the Administrators team, that is a group of project members with admin rights to module GitHub repositories.
Module owner: Each module shall have a module code owner. Module owners will have the overall responsibility of the contents of a Zephyr module repository. In particular, a module owner will:
coordinate code reviewing in the module repository
be the default assignee in pull-requests against the repository’s main branch
request additional collaborators to be added to the repository, as they see fit
regularly synchronize the module repository with its upstream counterpart following the policies described in Synchronizing with upstream
be aware of security vulnerability issues in the external project and update the module repository to include security fixes, as soon as the fixes are available in the upstream code base
list any known security vulnerability issues, present in the module codebase, in Zephyr release notes.
Note
Module owners are not required to be Zephyr Maintainers.
Merger: The Zephyr Release Engineering team has the right and the responsibility to merge approved pull requests in the main branch of a module repository.
Maintaining the module codebase¶
Updates in the zephyr main tree, for example, in public Zephyr APIs, may require patching a module’s codebase. The responsibility for keeping the module codebase up to date is shared between the contributor of such updates in Zephyr and the module owner. In particular:
the contributor of the original changes in Zephyr is required to submit the corresponding changes that are required in module repositories, to ensure that Zephyr CI on the pull request with the original changes, as well as the module integration testing are successful.
the module owner has the overall responsibility for synchronizing and testing the module codebase with the zephyr main tree. This includes occasional advanced testing of the module’s codebase in addition to the testing performed by Zephyr’s CI. The module owner is required to fix issues in the module’s codebase that have not been caught by Zephyr pull request CI runs.
Contributing changes to modules¶
Submitting and merging changes directly to a module’s codebase, that is, before they have been merged in the corresponding external project repository, should be limited to:
changes required due to updates in the zephyr main tree
urgent changes that should not wait to be merged in the external project first, such as fixes to security vulnerabilities.
Non-trivial changes to a module’s codebase, including changes in the module design or functionality should be discouraged, if the module has an upstream project repository. In that case, such changes shall be submitted to the upstream project, directly.
Submitting changes to modules describes in detail the process of contributing changes to module repositories.
Contribution guidelines¶
Contributing to Zephyr modules shall follow the generic project Contribution guidelines.
Pull Requests: may be merged with minimum of 2 approvals, including an approval by the PR assignee. In addition to this, pull requests in module repositories may only be merged if the introduced changes are verified with Zephyr CI tools, as described in more detail in other sections on this page.
The merging of pull requests in the main branch of a module repository must be coupled with the corresponding manifest file update in the zephyr main tree.
Issue Reporting: GitHub issues are intentionally disabled in module repositories, in favor of a centralized policy for issue reporting. Tickets concerning, for example, bugs or enhancements in modules shall be opened in the main zephyr repository. Issues should be appropriately labeled using GitHub labels corresponding to each module, where applicable.
Note
It is allowed to file bug reports for zephyr modules to track the corresponding upstream project bugs in Zephyr. These bug reports shall not affect the Release Quality Criteria.
Licensing requirements and policies¶
All source files in a module’s codebase shall include a license header, unless the module repository has main license file that covers source files that do not include license headers.
Main license files shall be added in the module’s codebase by Zephyr developers, only if they exist as part of the external project, and they contain a permissive OSI-compliant license. Main license files should preferably contain the full license text instead of including an SPDX license identifier. If multiple main license files are present it shall be made clear which license applies to each source file in a module’s codebase.
Individual license headers in module source files supersede the main license.
Any new content to be added in a module repository will require to have license coverage.
Note
Zephyr recommends conveying module licensing via individual license headers and main license files. This not a hard requirement; should an external project have its own practice of conveying how licensing applies in the module’s codebase (for example, by having a single or multiple main license files), this practice may be accepted by and be referred to in the Zephyr module, as long as licensing requirements, for example OSI compliance, are satisfied.
License policies¶
When creating a module repository a developer shall:
import the main license files, if they exist in the external project, and
document (for example in the module README or .yml file) the default license that covers the module’s codebase.
License checks¶
License checks (via CI tools) shall be enabled on every pull request that adds new content in module repositories.
Documentation requirements¶
All Zephyr module repositories shall include an .rst file documenting:
the scope and the purpose of the module
how the module integrates with Zephyr
the owner of the module repository
synchronization information with the external project (commit, SHA, version etc.)
licensing information as described in Licensing requirements and policies.
The file shall be required for the inclusion of the module and the contained information should be kept up to date.
Testing requirements¶
All Zephyr modules should provide some level of integration testing, ensuring that the integration with Zephyr works correctly. Integration tests:
may be in the form of a minimal set of samples and tests that reside in the zephyr main tree
should verify basic usage of the module (configuration, functional APIs, etc.) that is integrated with Zephyr.
shall be built and executed (for example in QEMU) as part of sanitycheck runs in pull requests that introduce changes in module repositories.
Note
New modules, that are candidates for being included in the Zephyr default manifest, shall provide some level of integration testing.
Note
Vendor HALs are implicitly tested via Zephyr tests built or executed on target platforms, so they do not need to provide integration tests.
The purpose of integration testing is not to provide functional verification of the module; this should be part of the testing framework of the external project.
Certain external projects provide test suites that reside in the upstream testing infrastructure but are written explicitly for Zephyr. These tests may (but are not required to) be part of the Zephyr test framework.
Deprecating and removing modules¶
Modules may be deprecated for reasons including, but not limited to:
Lack of maintainership in the module
Licensing changes in the external project
Codebase becoming obsolete
The module information shall indicate whether a module is deprecated and the build system shall issue a warning when trying to build Zephyr using a deprecated module.
Deprecated modules may be removed from the Zephyr default manifest after 2 Zephyr releases.
Note
Repositories of removed modules shall remain accessible via their original URL, as they are required by older Zephyr versions.
Integrate modules in Zephyr build system¶
The build system variable ZEPHYR_MODULES is a CMake list of
absolute paths to the directories containing Zephyr modules. These modules
contain CMakeLists.txt
and Kconfig
files describing how to
build and configure them, respectively. Module CMakeLists.txt
files are
added to the build using CMake’s add_subdirectory() command, and the
Kconfig
files are included in the build’s Kconfig menu tree.
If you have west installed, you don’t need to worry about how
this variable is defined unless you are adding a new module. The build system
knows how to use west to set ZEPHYR_MODULES. You can add additional
modules to this list by setting the ZEPHYR_EXTRA_MODULES CMake
variable or by adding a ZEPHYR_EXTRA_MODULES line to .zephyrrc
(See the section on Environment Variables for more details). This can be
useful if you want to keep the list of modules found with west and also add
your own.
See the section about Multiple Repository Management for more details.
Finally, you can also specify the list of modules yourself in various ways, or not use modules at all if your application doesn’t need them.
Module yaml file description¶
A module can be described using a file named zephyr/module.yml
.
The format of zephyr/module.yml
is described in the following:
Build files¶
Inclusion of build files, CMakeLists.txt
and Kconfig
, can be
described as:
build:
cmake: <cmake-directory>
kconfig: <directory>/Kconfig
The cmake: <cmake-directory>
part specifies that
<cmake-directory>
contains the CMakeLists.txt
to use. The
kconfig: <directory>/Kconfig
part specifies the Kconfig file to use.
Neither is required: cmake
defaults to zephyr
, and kconfig
defaults to zephyr/Kconfig
.
Here is an example module.yml
file referring to
CMakeLists.txt
and Kconfig
files in the root directory of the
module:
build:
cmake: .
kconfig: Kconfig
Build system integration¶
When a module has a module.yml
file, it will automatically be included into
the Zephyr build system. The path to the module is then accessible through Kconfig
and CMake variables.
In both Kconfig and CMake, the variable ZEPHYR_<module-name>_MODULE_DIR
contains the absolute path to the module.
In CMake, ZEPHYR_<module-name>_CMAKE_DIR
contains the
absolute path to the directory containing the CMakeLists.txt
file that
is included into CMake build system. This variable’s value is empty if the
module.yml file does not specify a CMakeLists.txt.
To read these variables for a Zephyr module named foo
:
In CMake: use
${ZEPHYR_FOO_MODULE_DIR}
for the module’s top level directory, and${ZEPHYR_FOO_CMAKE_DIR}
for the directory containing itsCMakeLists.txt
In Kconfig: use
$(ZEPHYR_FOO_MODULE_DIR)
for the module’s top level directory
Notice how a lowercase module name foo
is capitalized to FOO
in both CMake and Kconfig.
These variables can also be used to test whether a given module exists.
For example, to verify that foo
is the name of a Zephyr module:
if(ZEPHYR_FOO_MODULE_DIR)
# Do something if FOO exists.
endif()
In Kconfig, the variable may be used to find additional files to include.
For example, to include the file some/Kconfig
in module foo
:
source "$(ZEPHYR_FOO_MODULE_DIR)/some/Kconfig"
During CMake processing of each Zephyr module, the following two variables are also available:
the current module’s top level directory:
${ZEPHYR_CURRENT_MODULE_DIR}
the current module’s
CMakeLists.txt
directory:${ZEPHYR_CURRENT_CMAKE_DIR}
This removes the need for a Zephyr module to know its own name during CMake
processing. The module can source additional CMake files using these CURRENT
variables. For example:
include(${ZEPHYR_CURRENT_MODULE_DIR}/cmake/code.cmake)
Build settings¶
It is possible to specify additional build settings that must be used when including the module into the build system.
All root
settings are relative to the root of the module.
Build settings supported in the module.yml
file are:
board_root
: Contains additional boards that are available to the build system. Additional boards must be located in a<board_root>/boards
folder.dts_root
: Contains additional dts files related to the architecture/soc families. Additional dts files must be located in a<dts_root>/dts
folder.soc_root
: Contains additional SoCs that are available to the build system. Additional SoCs must be located in a<soc_root>/soc
folder.arch_root
: Contains additional architectures that are available to the build system. Additional architectures must be located in a<arch_root>/arch
folder.
Example of a module.yaml
file containing additional roots, and the
corresponding file system layout.
build:
settings:
board_root: .
dts_root: .
soc_root: .
arch_root: .
requires the following folder structure:
<module-root>
├── arch
├── boards
├── dts
└── soc
Sanitycheck¶
To execute both tests and samples available in modules, the Zephyr test runner
(sanitycheck) should be pointed to the directories containing those samples and
tests. This can be done by specifying the path to both samples and tests in the
zephyr/module.yml
file. Additionally, if a module defines out of tree
boards, the module file can point sanitycheck to the path where those files
are maintained in the module. For example:
build:
cmake: .
samples:
- samples
tests:
- tests
boards:
- boards
Module Inclusion¶
Using West¶
If west is installed and ZEPHYR_MODULES is not already set, the build system finds all the modules in your west installation and uses those. It does this by running west list to get the paths of all the projects in the installation, then filters the results to just those projects which have the necessary module metadata files.
Each project in the west list
output is tested like this:
If the project contains a file named
zephyr/module.yml
, then the content of that file will be used to determine which files should be added to the build, as described in the previous section.Otherwise (i.e. if the project has no
zephyr/module.yml
), the build system looks forzephyr/CMakeLists.txt
andzephyr/Kconfig
files in the project. If both are present, the project is considered a module, and those files will be added to the build.If neither of those checks succeed, the project is not considered a module, and is not added to ZEPHYR_MODULES.
Without West¶
If you don’t have west installed or don’t want the build system to use it to
find Zephyr modules, you can set ZEPHYR_MODULES yourself using one
of the following options. Each of the directories in the list must contain
either a zephyr/module.yml
file or the files
zephyr/CMakeLists.txt
and Kconfig
, as described in the previous
section.
At the CMake command line, like this:
cmake -DZEPHYR_MODULES=<path-to-module1>[;<path-to-module2>[...]] ...
At the top of your application’s top level
CMakeLists.txt
, like this:set(ZEPHYR_MODULES <path-to-module1> <path-to-module2> [...]) find_package(Zephyr REQUIRED HINTS $ENV{ZEPHYR_BASE})
If you choose this option, make sure to set the variable before calling
find_package(Zephyr ...)
, as shown above.In a separate CMake script which is pre-loaded to populate the CMake cache, like this:
# Put this in a file with a name like "zephyr-modules.cmake" set(ZEPHYR_MODULES <path-to-module1> <path-to-module2> CACHE STRING "pre-cached modules")
You can tell the build system to use this file by adding
-C zephyr-modules.cmake
to your CMake command line.
Not using modules¶
If you don’t have west installed and don’t specify ZEPHYR_MODULES yourself, then no additional modules are added to the build. You will still be able to build any applications that don’t require code or Kconfig options defined in an external repository.
Submitting changes to modules¶
When submitting new or making changes to existing modules the main repository Zephyr needs a reference to the changes to be able to verify the changes. In the main tree this is done using revisions. For code that is already merged and part of the tree we use the commit hash, a tag, or a branch name. For pull requests however, we require specifying the pull request number in the revision field to allow building the zephyr main tree with the changes submitted to the module.
To avoid merging changes to master with pull request information, the pull
request should be marked as DNM
(Do Not Merge) or preferably a draft pull
request to make sure it is not merged by mistake and to allow for the module to
be merged first and be assigned a permanent commit hash. Once the module is
merged, the revision will need to be changed either by the submitter or by the
maintainer to the commit hash of the module which reflects the changes.
Note that multiple and dependent changes to different modules can be submitted using exactly the same process. In this case you will change multiple entries of all modules that have a pull request against them.
Process for submitting a new module¶
A request for a new module should be initialized using an RFC issue in the Zephyr project issue tracking system with details about the module and how it integrates into the project. If the request is approved, a new repository will created by the project team and initialized with basic information that would allow submitting code to the module project following the project contribution guidelines.
If a module is maintained as a fork of another project on Github, the Zephyr module
related files and changes in relation to upstream need to be maintained in a
special branch named zephyr
.
Follow the following steps to request a new module:
Use GitHub issues to open an issue with details about the module to be created
Propose a name for the repository to be created under the Zephyr project organization on Github.
Maintainers from the Zephyr project will create the repository and initialize it. You will be added as a collaborator in the new repository.
Submit the module content (code) to the new repository following the guidelines described here.
Add a new entry to the west.yml with the following information:
- name: <name of repository> path: <path to where the repository should be cloned> revision: <ref pointer to module pull request>
For example, to add my_module to the manifest:
- name: my_module
path: modules/lib/my_module
revision: pull/23/head
Where 23 in the example above indicated the pull request number submitted to the my_module repository. Once the module changes are reviewed and merged, the revision needs to be changed to the commit hash from the module repository.
Process for submitting changes to existing modules¶
Submit the changes using a pull request to an existing repository following the contribution guidelines.
Submit a pull request changing the entry referencing the module into the west.yml of the main Zephyr tree with the following information:
- name: <name of repository> path: <path to where the repository should be cloned> revision: <ref pointer to module pull request>
For example, to add my_module to the manifest:
- name: my_module
path: modules/lib/my_module
revision: pull/23/head
Where 23 in the example above indicated the pull request number submitted to the my_module repository. Once the module changes are reviewed and merged, the revision needs to be changed to the commit hash from the module repository.