Inglorious kernel developer workflow tool


Rodrigo Siqueira <>


Matheus Tavares <>





Manual section


Manual group

Kernel Workflow


kw <command> [<option> …]


kw mission is: reduce the overhead related with infrastructure project setup in projects that have a similar workflow to the Linux Kernel. It can (and should) be customized by editing the kworkflow.config file, as discussed in section ABOUT kworflow.config.


kw offers several subcommands catering to different audiences and targeting different functionality groups. Most subcommands have sane defaults.


When we develop for Linux Kernel, we continuously want to install/update the current version of the Kernel image or modules, and this tasks may require several steps to be accomplished. For this reason, kw provides an option named deploy that handles attempts to handle all the complexity related to the new Kernel installation. It is essential to highlight that we try to support three different types of deploy: local and remote. When you want to update your host machine, you can use the local option. Finally, we provide the remote option, which is much more flexible since it uses network; notice that this approach is the most generic one because you can use it for vm and local.


Projects that have a similar workflow to the Linux Kernel usually have a set of tools that simplify part of the tasks related with the code. This section describes some of the key features supported by kw to help with code.


Linux kernel has multiple subsystems that expose operations via sysfs or provide mechanisms for userspace to interact with the driver. For this reason, kw offers some options that target some specific subsystems for providing facilities for users to interact with a particular subsystem. Currently, we only support drm.


Linux kernel provides multiple mechanisms for debugging; in particular, kw tries to simplify the debug process for three of them: events, ftrace, and dmesg. All the debug options are intended to support remote and local targets.


This section describes a tool available in kw to help developers keep track of configuration files and other features provided by kw that do not fit in the previous sections.


Clean all files generated by kw.

h, help

Show basic help.

version, –version, -v

Show kworkflow version.

ABOUT kworflow.config

kw reads its configuration from two files: the global <path>/etc/kworkflow.config file and the local kworkflow.config file present at the current working directory. The global kworkflow.config is a part of the kw code and provides the overall behavior for kw. Local kworkflow.config settings override global ones; you may have one kworkflow.config per project. In this section, we describe the possible fields you can specify in the configuration files.


Sets the user to be used by ssh. By default kw uses root.


Sets the IP address to be used by ssh. By default kw uses localhost.


Sets the ssh port. By default kw uses 2222.


Provides an optional SSH configuration file to be used by ssh. For more details see man ssh_config.


Sets the hostname to be used when an SSH configuration file is provided.


Allows you to specify the default architecture used by kw. By default, kw uses x86_64.


Use this option as a way to indicate to kw the kernel image name. This is the file present in the directory arch/*/boot/; keep in mind that the kernel image name might change based on the user config file or target architecture.


Kw supports cross compile setup, use this option to indicate the target toolchain.

alert=[vs | s | v | n]

Default alert options, you have:

  1. v: enables visual notification.

  2. s: enables sound notification.

  3. vs or sv: enables both.

  4. n (or any other option): disables notifications.


Command to run for sound completion alert. By default, kw uses paplay INSTALLPATH/sounds/complete.wav &


Command to run for visual completion alert. By default, kw uses notify-send -i checkbox -t 10000 "kw" "Command: \\"$COMMAND\\" completed!"


You may use the COMMAND variable, which will be replaced by the kw command whose conclusion the user wished to be alerted of.


By default, kw deploys in the remote; however, you can change this behavior with this variable. The available options are: local and remote.


Reboot machine after the deploy finishes.


This option is disabled by default, if enabled, it requires a command that instructs kw to turn on the GUI.


This option is disabled by default, if enabled, it requires a command that instructs kw to turn off the GUI.


For these examples, we suppose the fields in your kworkflow.config file is already configured.

First, if you are working in a specific kernel module, and if you want to install your recent changes in your local machine you can use:

cd <kernel-path>
kw d --local --modules

For building and installing a new module version based on the current kernel version, you can use:

cd <kernel-path>
kw bd

For checking the code style:

cd <kernel-path>
kw c drivers/iio/dummy/
kw c drivers/iio/dummy/iio_simple_dummy.c

If you want to check the maintainers:

cd <kernel-path>
kw m drivers/iio/dummy/iio_simple_dummy.c

In case you want that kw saves your current .config file, you can use:

cd <kernel-path>
kw k --save my_current_config

You can see the config’s file maintained by kw with:

kw k --list

You can turn on your VM with:

kw u

After you start your VM you can ssh into it with:

kw s -c="dmesg -wH"
kw s

You can see data related to your kw usage by using the --statistics flag on the report option, see some examples below:

kw report --statistics --day
kw report --statistics --week
kw report --statistics --month
kw report --statistics --year

You can also request a specific day, week, month, or year. For example:

kw report --statistics --day=2020/05/12
kw report --statistics --week=2020/02/29
kw report --statistics --month=2020/04
kw report --statistics --year=1984

If you are working with DRM drivers, you can take advantage of load and unload commands combined with GUI control commands. For example:

kw drm --load-module='amdgpu' --gui-on # Load a driver and trigger the user GUI
kw drm --unload-module='amdgpu' # Turn off user GUI and unload the driver

If you need to debug an issue based on event values, you can try the debug options. For example:

kw debug --list # Show all events debug available in the target
kw debug --list --event="amdgpu_dm" # Show all events available under amdgpu_dm
kw debug --event='amdgpu_dm:amdgpu_dm_dce_clocks_state[sclk_khz > 0]' # Enable amdgpu_dm_dce_clocks_state event and filter by sclk_khz > 0
kw debug --disable --event='amdgpu_dm:amdgpu_dm_dce_clocks_state' # Disable amdgpu_dm_dce_clocks_state events
kw debug --event='amdgpu_dm:amdgpu_dm_dce_clocks_state' --history # Save each debug in a separated set of files
kw debug --event='amdgpu_dm:amdgpu_dm_dce_clocks_state' --follow # Wait for new event message
kw debug --event='amdgpu_dm:amdgpu_dm_dce_clocks_state' --cmd="export DISPLAY=:0.0 && xrandr --props" # Enable amdgpu_dm_dce_clocks_state, run "export DISPLAY=:0.0 && xrandr --props", collect logs, and disable events


You have to wait for the sshd to become ready.