This covers how to install the toolchain on your platform to start using and developing Tock.
- Requirements
- Compiling the Kernel
- Loading the kernel onto a board
- Installing your first application
- Compiling applications
- Developing TockOS
-
rustup to install Rust (version >= 1.11.0)
-
Host toolchain (gcc, glibc)
-
Command line utilities: make, find
-
A supported board or QEMU configuration.
If you are just starting to work with TockOS, you should look in the
boards/
subdirectory and choose one of the options withtockloader
support to load applications, as that is the configuration that most examples and tutorials assume.Note: QEMU support in Tock is in the early stages. Please be sure to check whether and how QEMU is supported for a board based on the table in the
boards/
subdirectory. Themake ci-job-qemu
target is the authority on QEMU support.- Info about testing Tock on QEMU
- 01/08/2020 : Among the boards supported by Tock, SiFive HiFive1 RISC-V Board can be tested in QEMU.
- Info about testing Tock on QEMU
Nix:
$ nix-shell
MacOS:
$ curl https://sh.rustup.rs -sSf | sh
$ pip3 install --upgrade tockloader
Ubuntu:
$ apt install -y build-essential python3-pip curl
$ curl https://sh.rustup.rs -sSf | sh
$ pip3 install --upgrade tockloader --user
$ grep -q dialout <(groups $(whoami)) || sudo usermod -a -G dialout $(whoami) # Note, will need to reboot if prompted for password
Then build the kernel by running make
in the boards/<platform>
directory.
These steps go into a little more depth. Note that the build system is capable of installing some of these tools, but you can also install them yourself.
We are using nightly-2023-04-13
. We require
installing it with rustup so you can manage multiple
versions of Rust and continue using stable versions for other Rust code:
$ curl https://sh.rustup.rs -sSf | sh
This will install rustup
in your home directory, so you will need to
source ~/.profile
or open a new shell to add the .cargo/bin
directory
to your $PATH
.
Then install the correct nightly version of Rust:
$ rustup install nightly-2023-04-13
tockloader
programs the kernel and applications onto boards, and also has
features that are generally useful for all Tock boards, such as easy-to-manage
serial connections, along with the ability to list, add, replace, and remove
applications over JTAG (or USB if a bootloader is installed).
- tockloader (version >= 1.0)
Tockloader is a Python application and can be installed with the Python package manager (pip).
(Linux): pip3 install --upgrade tockloader --user
(MacOS): pip3 install --upgrade tockloader
Tock builds a unique kernel for every board it supports. Boards include
details like pulling together the correct chips and pin assignments. To
build a kernel, first choose a board, then navigate to that board directory.
e.g. cd boards/nordic/nrf52840dk ; make
.
Some boards have special build options that can only be used within the board's directory. All boards share a few common targets:
all
(default): Compile Tock for this board.debug
: Generate build(s) for debugging support, details vary per board.doc
: Build documentation for this board.clean
: Remove built artifacts for this board.flash
: Load code using JTAG, if available.program
: Load code using a bootloader, if available.
The board-specific READMEs in each board's subdirectory provide more details for each platform.
The process to load the kernel onto the board depends on the board. You should be able to program the kernel by running:
$ make install
For some boards, you will need a programming adapter to flash code. Depending on
the board and which adapter it supports, you will need either the free openocd
or Segger's proprietary JLinkExe
. Programming adapters are available as
standalone devices (for example the JLink EDU JTAG
debugger available on
Digikey),
but most development boards come with an onboard programming and debugging
adapter. In that case, the board you use determines which software you will need
and the Makefile
in the board directory will know which one to call. Again,
the board-specific READMEs provide the required details.
JLink
is available from the Segger
website. You want to install
the "J-Link Software and Documentation Pack". There are various
packages available depending on operating system. We require a version
greater than or equal to 5.0
.
Openocd
works with various programming and debugging adapters. For
most purposes, available distribution packages are sufficient and it can
be installed with:
(Linux/Debian): sudo apt-get install openocd
(MacOS): brew install open-ocd
We require at least version 0.8.0
to support the SAM4L on imix
if
you choose to flash it using an adapter instead of the bootloader.
Some boards (at the time of writing the HiFive1 RISC-V board) may
require newer or unreleased versions, in that case you should follow
the installation instructions on the openocd
website.
Depending on which programming adapter you use, you may want to add a
udev
rule in /etc/udev/rules.d
that allows you to interact with
the board as a user instead of as root. If you install the deb
packet of the JLink
software it will automatically install a
/etc/udev/rules.d/99-jlink.rules
that allows everyone to access the
adapter. If you use something else, like for example the onboard
programmer of a ST Nucleo board, you could install something like this as
/etc/udev/rules.d/99-stlinkv2-1.rules
:
# stm32 nucleo boards, with onboard st/linkv2-1
# ie, STM32F0, STM32F4.
# STM32VL has st/linkv1, which is quite different
SUBSYSTEMS=="usb", ATTRS{idVendor}=="0483", ATTRS{idProduct}=="374b", \
MODE:="0660", GROUP="dialout", \
SYMLINK+="stlinkv2-1_%n"
A kernel alone isn't much use, as an embedded developer you want to
see some LEDs blink. Fortunately, there is an example blink
app
available from the TockOS app repository which tockloader
can
download and install (if you are using a board that is supported
by tockloader
).
For certain boards (e.g. Hail and imix), tockloader
can read
attributes from the board to configure how it communicates with the
board. For many boards, however, tockloader
cannot know
which board and communication method you want to use, so you have to
tell it explicitly. For example:
$ tockloader install --board nrf52dk --jlink blink
Could not find TAB named "blink" locally.
[0] No
[1] Yes
Would you like to check the online TAB repository for that app?[0] 1
Installing apps on the board...
Using known arch and jtag-device for known board nrf52dk
Finished in 2.567 seconds
Boards that use openocd
will of course require the parameter
--openocd
instead of --jlink
. If your board has a serial
bootloader, tockloader
should work without any additional arguments:
$ tockloader install blink
However, you can specify the board type manually as well:
$ tockloader install --board imix blink
You can also tell it which serial port to use (which is useful if you
have multiple boards plugged in) by passing it the --port
parameter
like --port /dev/ttyACM0
to use /dev/ttyACM0
:
$ tockloader --port /dev/ttyACM0 install blink
To see the list of boards tockloader
knows about you can run:
$ tockloader list-known-boards
If everything has worked until here, the LEDs on your board should now display a binary counter. Congratulations, you have a working TockOS installation on your board!
The last remaining step is to compile applications locally. All user-level code lives in two separate repositories:
- libtock-c: C and C++ apps.
- libtock-rs: Rust apps.
The C version of the Tock library and the example applications is older and more stable, so it is a good idea to look at these first. So look at the libtock-c README and follow the steps therein. Then you can do the same for the libtock-rs README. This should give you a first impression of how to build and deploy applications for TockOS.
For an introduction on how applications work in TockOS, have a look at the "Userland" document in this directory.
Rust includes a tool for automatically formatting Rust source code. Simply run:
$ make format
from the root of the repository to format all rust code in the repository.
Occasionally, Tock updates to a new nightly version of Rust. The build system
automatically checks whether the versions of rustc
and rustup
are correct
for the build requirements, and updates them when necessary. After the
installation of the initial four requirements, you shouldn't have to worry
about keeping them up to date.