Embedded Linux system development

Embedded Linux system

development

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
Latest update: July 19, 2020.

Document updates and sources:

https://bootlin.com/doc/training/embedded- linux

Corrections, suggestions, contributions and translations are welcome! embedded Linux and kernel engineering
Send them to [email protected]

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 1/520
 Rights to copy

© Copyright 2004-2020, Bootlin

License: Creative Commons Attribution - Share Alike 3.0
https://creativecommons.org/licenses/by-sa/3.0/legalcode
You are free:
▶ to copy, distribute, display, and perform the work
▶ to make derivative works
▶ to make commercial use of the work
Under the following conditions:
▶ Attribution. You must give the original author credit.
▶ Share Alike. If you alter, transform, or build upon this work, you may distribute the resulting work only
under a license identical to this one.
▶ For any reuse or distribution, you must make clear to others the license terms of this work.
▶ Any of these conditions can be waived if you get permission from the copyright holder.
Your fair use and other rights are in no way affected by the above.

Document sources: https://github.com/bootlin/training-materials/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 2/520
 Hyperlinks in the document

There are many hyperlinks in the document

▶ Regular hyperlinks:
https://kernel.org/
▶ Kernel documentation links:
dev-tools/kasan
▶ Links to kernel source files and directories:
drivers/input/
include/linux/fb.h
▶ Links to the declarations, definitions and instances of kernel symbols (functions,
types, data, structures):
platform_get_irq()
GFP_KERNEL
struct file_operations

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 3/520
 Company at a glance

▶ Engineering company created in 2004, named ”Free Electrons” until Feb. 2018.
▶ Locations: Orange, Toulouse, Lyon (France)
▶ Serving customers all around the world
▶ Head count: 12 - Only Free Software enthusiasts!
▶ Focus: Embedded Linux, Linux kernel, build systems and low level Free and Open
Source Software for embedded and real-time systems.
▶ Bootlin is often in the top 20 companies contributing to the Linux kernel.
▶ Activities: development, training, consulting, technical support.
▶ Added value: get the best of the user and development community and the
resources it offers.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 4/520
 Bootlin on-line resources

▶ All our training materials and technical

presentations:
https://bootlin.com/docs/
▶ Technical blog:
https://bootlin.com/
▶ Quick news (Mastodon):
https://fosstodon.org/@bootlin
▶ Quick news (Twitter):
https://twitter.com/bootlincom
▶ Quick news (LinkedIn): Mastodon is a free and decentralized
https: social network created in the best
//www.linkedin.com/company/bootlin interests of its users.
▶ Elixir - browse Linux kernel sources on-line: Image credits: Jin Nguyen - https://frama.link/bQwcWHTP

https://elixir.bootlin.com
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 5/520
 Generic course information

Generic course
information

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 6/520
 Labs on QEMU virtual hardware

Practical labs on on the QEMU emulated ARM

Versatile Express Cortex A9 board
▶ Most features as on a real embedded board

EMU
▶ Can boot the Linux kernel from U-Boot
▶ Supports mainline Linux and U-Boot
▶ Networking possible with the emulated board
▶ Emulated SD card

Lab instructions available on

https://bootlin.com/doc/training/embedded-linux-qemu/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 7/520
 Labs proposed on another platform

After this course, you can also run all labs on the
Microchip SAMA5D3 Xplained ARM board.
In addition, you will also have real-time and NAND
flash labs!

Lab instructions available on https://bootlin.com/doc/training/embedded-linux/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 8/520
 Labs proposed on another platform

After this course, you can also run most labs on the
STM32MP157A-DK1 Discovery board
(https://www.st.com/en/evaluation-
tools/stm32mp157a-dk1.html)

Lab instructions available on

https://bootlin.com/doc/training/embedded-linux-4d/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 9/520
 Participate!

During the lectures...

▶ Don’t hesitate to ask questions. Other people in the audience may have similar
questions too.
▶ This helps the trainer to detect any explanation that wasn’t clear or detailed
enough.
▶ Don’t hesitate to share your experience, for example to compare Linux with other
operating systems used in your company.
▶ Your point of view is most valuable, because it can be similar to your colleagues’
and different from the trainer’s.
▶ Your participation can make our session more interactive and make the topics
easier to learn.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 10/520
 Practical lab guidelines

During practical labs...

▶ We cannot support more than 8 workstations at once (each with its board and
equipment). Having more would make the whole class progress slower,
compromising the coverage of the whole training agenda (exception for public
sessions: up to 10 people).
▶ So, if you are more than 8 participants, please form up to 8 working groups.
▶ Open the electronic copy of your lecture materials, and use it throughout the
practical labs to find the slides you need again.
▶ Don’t hesitate to copy and paste commands from the PDF slides and labs.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 11/520
 Advise: write down your commands!

During practical labs, write down all your commands in a text file.
▶ You can save a lot of time re-using commands Lab commands

in later labs. Cross-compiling kernel:

export ARCH=arm
▶ This helps to replay your work if you make export CROSS_COMPILE=arm-linux-
make sama5_defconfig
significant mistakes.
Booting kernel through tftp:
▶ You build a reference to remember commands setenv bootargs console=ttyS0 root=/dev/nfs
setenv bootcmd tftp 0x21000000 zImage; tftp
in the long run. 0x22000000 dtb; bootz 0x21000000 - 0x2200...

▶ That’s particular useful to keep kernel Making ubifs images:

mkfs.ubifs -d rootfs -o root.ubifs -e 124KiB
command line settings that you used earlier. -m 2048 -c 1024

▶ Also useful to get help from the instructor, Encountered issues:

Restart NFS server after editing /etc/exports!
showing the commands that you run.
gedit ~/lab-history.txt

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 12/520
 Cooperate!

As in the Free Software and Open Source community, cooperation during practical labs
is valuable in this training session:
▶ If you complete your labs before other people, don’t hesitate to help other people
and investigate the issues they face. The faster we progress as a group, the more
time we have to explore extra topics.
▶ Explain what you understood to other participants when needed. It also helps to
consolidate your knowledge.
▶ Don’t hesitate to report potential bugs to your instructor.
▶ Don’t hesitate to look for solutions on the Internet as well.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 13/520
 Command memento sheet

▶ This memento sheet gives command examples

for the most typical needs (looking for files,
extracting a tar archive...)
▶ It saves us 1 day of UNIX / Linux command
line training.
▶ Our best tip: in the command line shell, always
hit the Tab key to complete command names
and file paths. This avoids 95% of typing
mistakes.
▶ Get an electronic copy on
https://bootlin.com/doc/legacy/command-
line/command_memento.pdf

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 14/520
 vi basic commands

▶ The vi editor is very useful to make quick

changes to files in an embedded target.
▶ Though not very user friendly at first, vi is
very powerful and its main 15 commands are
easy to learn and are sufficient for 99% of
everyone’s needs!
▶ Get an electronic copy on
https://bootlin.com/doc/legacy/command-
line/vi_memento.pdf
▶ You can also take the quick tutorial by running
vimtutor. This is a worthy investment!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 15/520
 Introduction to Embedded Linux

Introduction to
Embedded Linux

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 16/520
 Birth of Free Software

▶ 1983, Richard Stallman, GNU project and the free

software concept. Beginning of the development of gcc,
gdb, glibc and other important tools
▶ 1991, Linus Torvalds, Linux kernel project, a UNIX-like
operating system kernel. Together with GNU software and
many other open-source components: a completely free
operating system, GNU/Linux
▶ 1995, Linux is more and more popular on server systems
▶ 2000, Linux is more and more popular on embedded
systems Richard Stallman in 2019
▶ 2008, Linux is more and more popular on mobile devices Image credits (Wikipedia):
https://frama.link/qC73jkk4

▶ 2010, Linux is more and more popular on phones

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 17/520
 Free software?

▶ A program is considered free when its license offers to all its users the following
four freedoms
▶ Freedom to run the software for any purpose
▶ Freedom to study the software and to change it
▶ Freedom to redistribute copies
▶ Freedom to distribute copies of modified versions
▶ These freedoms are granted for both commercial and non-commercial use
▶ They imply the availability of source code, software can be modified and
distributed to customers
▶ Good match for embedded systems!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 18/520
 What is embedded Linux?

Embedded Linux is the usage of the Linux

kernel and various open-source components
in embedded systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 19/520
 Introduction to Embedded Linux

Advantages of Linux and open-source for

embedded systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 20/520
 Re-using components

▶ The key advantage of Linux and open-source in embedded systems is the ability
to re-use components
▶ The open-source ecosystem already provides many components for standard
features, from hardware support to network protocols, going through multimedia,
graphic, cryptographic libraries, etc.
▶ As soon as a hardware device, or a protocol, or a feature is wide-spread enough,
high chance of having open-source components that support it.
▶ Allows to quickly design and develop complicated products, based on existing
components.
▶ No-one should re-develop yet another operating system kernel, TCP/IP stack,
USB stack or another graphical toolkit library.
▶ Allows to focus on the added value of your product.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 21/520
 Low cost

▶ Free software can be duplicated on as many devices as you want, free of charge.
▶ If your embedded system uses only free software, you can reduce the cost of
software licenses to zero. Even the development tools are free, unless you choose
a commercial embedded Linux edition.
▶ Of course, using Linux is not free of cost. You still need substantial learning and
engineering efforts to achieve your goals.
▶ Allows to have a higher budget for the hardware or to increase the
company’s skills and knowledge

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 22/520
 Full control

▶ With open-source, you have the source code for all components in your system
▶ Allows unlimited modifications, changes, tuning, debugging, optimization, for an
unlimited period of time
▶ Without lock-in or dependency from a third-party vendor
▶ To be true, non open-source components must be avoided when the system is
designed and developed
▶ Allows to have full control over the software part of your system and
secure your investment

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 23/520
 Quality

▶ Many open-source components are widely used, on millions of systems

▶ Usually higher quality than what an in-house development can produce, or even
proprietary vendors
▶ Of course, not all open-source components are of good quality, but most of the
widely-used ones are.
▶ Allows to design your system with high-quality components at the
foundations

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 24/520
 Eases testing of new features

▶ Open-source being freely available, it is easy to get a piece of software and

evaluate it
▶ Allows to easily study several options while making a choice
▶ Much easier than purchasing and demonstration procedures needed with most
proprietary products
▶ Allows to easily explore new possibilities and solutions

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 25/520
 Community support

▶ Open-source software components are developed by communities of developers

and users
▶ This community can provide high-quality support: you can directly contact the
main developers of the component you are using. The likelyhood of getting an
answer doesn’t depend what company you work for.
▶ Often better than traditional support, but one needs to understand how the
community works to properly use the community support possibilities
▶ Allows to speed up the resolution of problems when developing your system

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 26/520
 Taking part into the community

▶ Possibility of taking part into the development community of some of the

components used in the embedded systems: bug reporting, test of new versions or
features, patches that fix bugs or add new features, etc.
▶ Most of the time the open-source components are not the core value of the
product: it’s the interest of everybody to contribute back.
▶ For the engineers: a very motivating way of being recognized outside the
company, communication with others in the same field, opening of new
possibilities, etc.
▶ For the managers: motivation factor for engineers, allows the company to be
recognized in the open-source community and therefore get support more easily
and be more attractive to open-source developers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 27/520
 Introduction to Embedded Linux

A few examples of embedded systems running

Linux

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 28/520
 Wireless routers

Image credits: Evan Amos (https://bit.ly/2JzDIkv)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 29/520
 Video systems

Image credits: https://bit.ly/2HbwyVq

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 30/520
 Bike computers

Product from BLOKS (http://bloks.de).

Permission to use this picture only in this document, in updates and in translations.
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 31/520
 Robots

eduMIP robot (https://www.ucsdrobotics.org/edumip)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 32/520
 Viticulture machine

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 33/520
 Introduction to Embedded Linux

Embedded hardware for Linux systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 34/520
 Processor and architecture (1)
The Linux kernel and most other architecture-dependent components support a wide
range of 32 and 64 bits architectures
▶ x86 and x86-64, as found on PC platforms, but also embedded systems
(multimedia, industrial)
▶ ARM, with hundreds of different SoCs (all sorts of products)
▶ RiscV, the rising architecture with a free instruction set (from high-end cloud
computing to the smallest embedded systems)
▶ PowerPC (mainly real-time, industrial applications)
▶ MIPS (mainly networking applications)
▶ SuperH (mainly set top box and multimedia applications)
▶ c6x (TI DSP architecture)
▶ Microblaze (soft-core for Xilinx FPGA)
▶ Others: ARC, m68k, Xtensa...
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 35/520
 Processor and architecture (2)

▶ Both MMU and no-MMU architectures are supported, even though no-MMU
architectures have a few limitations.
▶ Linux does not support small microcontrollers (8 or 16 bit)
▶ Besides the toolchain, the bootloader and the kernel, all other components are
generally architecture-independent

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 36/520
 RAM and storage

▶ RAM: a very basic Linux system can work within 8 MB of RAM, but a more
realistic system will usually require at least 32 MB of RAM. Depends on the type
and size of applications.
▶ Storage: a very basic Linux system can work within 4 MB of storage, but usually
more is needed.
▶ Flash storage is supported, both NAND and NOR flash, with specific filesystems
▶ Block storage including SD/MMC cards and eMMC is supported
▶ Not necessarily interesting to be too restrictive on the amount of RAM/storage:
having flexibility at this level allows to re-use as many existing components as
possible.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 37/520
 Communication

▶ The Linux kernel has support for many common communication buses
▶ I2C
▶ SPI
▶ CAN
▶ 1-wire
▶ SDIO
▶ USB
▶ And also extensive networking support
▶ Ethernet, Wifi, Bluetooth, CAN, etc.
▶ IPv4, IPv6, TCP, UDP, SCTP, DCCP, etc.
▶ Firewalling, advanced routing, multicast

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 38/520
 Types of hardware platforms (1)

▶ Evaluation platforms from the SoC vendor. Usually

expensive, but many peripherals are built-in. Generally
unsuitable for real products. STM32MP157C-EV1
evaluation board
▶ Component on Module, a small board with only Image credits (st.com):
https://frama.link/cFseMy8H
CPU/RAM/flash and a few other core components, with
connectors to access all other peripherals. Can be used to
build end products for small to medium quantities.

PocketBeagle
Image credits (Beagleboard.org):
https://beagleboard.org/pocket

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 39/520
 Types of hardware platforms (2)

▶ Community development platforms, to make a particular SoC popular and

easily available. These are ready-to-use and low cost, but usually have less
peripherals than evaluation platforms. To some extent, can also be used for real
products.
▶ Custom platform. Schematics for evaluation boards or development platforms
are more and more commonly freely available, making it easier to develop custom
platforms.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 40/520
 Criteria for choosing the hardware

▶ Make sure the hardware you plan to use is already supported by the Linux kernel,
and has an open-source bootloader, especially the SoC you’re targeting.
▶ Having support in the official versions of the projects (kernel, bootloader) is a lot
better: quality is better, and new versions are available.
▶ Some SoC vendors and/or board vendors do not contribute their changes back to
the mainline Linux kernel. Ask them to do so, or use another product if you can.
A good measurement is to see the delta between their kernel and the official one.
▶ Between properly supported hardware in the official Linux kernel and
poorly-supported hardware, there will be huge differences in development
time and cost.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 41/520
 Introduction to Embedded Linux

Embedded Linux system architecture

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 42/520
 Host and target

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 43/520
 Software components

▶ Cross-compilation toolchain
▶ Compiler that runs on the development machine, but generates code for the target
▶ Bootloader
▶ Started by the hardware, responsible for basic initialization, loading and executing
the kernel
▶ Linux Kernel
▶ Contains the process and memory management, network stack, device drivers and
provides services to user space applications
▶ C library
▶ The interface between the kernel and the user space applications
▶ Libraries and applications
▶ Third-party or in-house

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 44/520
 Embedded Linux work

Several distinct tasks are needed when deploying embedded Linux in a product:
▶ Board Support Package development
▶ A BSP contains a bootloader and kernel with the suitable device drivers for the
targeted hardware
▶ Purpose of our Kernel Development training
▶ System integration
▶ Integrate all the components, bootloader, kernel, third-party libraries and
applications and in-house applications into a working system
▶ Purpose of this training
▶ Development of applications
▶ Normal Linux applications, but using specifically chosen libraries

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 45/520
 Embedded Linux development environment

Embedded Linux
development
environment

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 46/520
 Embedded Linux solutions

▶ Two ways to switch to embedded Linux

▶ Use solutions provided and supported by vendors like MontaVista, Wind River or
TimeSys. These solutions come with their own development tools and environment.
They use a mix of open-source components and proprietary tools.
▶ Use community solutions. They are completely open, supported by the community.
▶ In Bootlin training sessions, we do not promote a particular vendor, and therefore
use community solutions
▶ However, knowing the concepts, switching to vendor solutions will be easy

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 47/520
 OS for Linux development
We strongly recommend to use GNU/Linux as the desktop operating system to
embedded Linux developers, for multiple reasons.
▶ All community tools are developed and designed to run on Linux. Trying to use
them on other operating systems (Windows, Mac OS X) will lead to trouble, and
their usage on these systems is generally not supported by community developers.
▶ As Linux also runs on the embedded device, all the knowledge gained from using
Linux on the desktop will apply similarly to the embedded device.
▶ If you are stuck with a Windows desktop, at least you should use GNU/Linux in a
virtual machine (such as VirtualBox which is open source), though there could be
a small performance penalty.

:-)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 48/520
 Desktop Linux distribution

▶ Any good and sufficiently recent Linux desktop

distribution can be used for the development
workstation
▶ Ubuntu, Debian, Fedora, openSUSE, Red Hat, etc.
▶ We have chosen Ubuntu, as it is a widely used and
easy to use desktop Linux distribution
▶ The Ubuntu setup on the training laptops has
intentionally been left untouched after the normal
installation process. Learning embedded Linux is also
about learning the tools needed on the development
workstation!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 49/520
 Linux root and non-root users

▶ Linux is a multi-user operating system

▶ The root user is the administrator, and it can do privileged operations such as:
mounting filesystems, configuring the network, creating device files, changing the
system configuration, installing or removing software
▶ All other users are unprivileged, and cannot perform these administrator-level
operations
▶ On an Ubuntu system, it is not possible to log in as root, only as a normal user.
▶ The system has been configured so that the user account created first is allowed
to run privileged operations through a program called sudo.
▶ Example: sudo mount /dev/sda2 /mnt/disk

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 50/520
 Software packages

▶ The distribution mechanism for software in GNU/Linux is different from the one
in Windows
▶ Linux distributions provides a central and coherent way of installing, updating and
removing applications and libraries: packages
▶ Packages contains the application or library files, and associated
meta-information, such as the version and the dependencies
▶ .deb on Debian and Ubuntu, .rpm on Red Hat, Fedora, openSUSE
▶ Packages are stored in repositories, usually on HTTP or FTP servers
▶ You should only use packages from official repositories for your distribution, unless
strictly required.
▶ Note: Snap and Flatpak offer new ways of packaging applications in a
self-contained way. See https://www.atechtown.com/flatpak-vs-snap/.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 51/520
 Managing software packages (1)

Instructions for Debian based GNU/Linux systems

(Debian, Ubuntu...)
▶ Package repositories are specified in /etc/apt/sources.list
▶ To update package repository lists:
sudo apt update
▶ To find the name of a package to install, the best is to use the search engine on
https://packages.debian.org or on https://packages.ubuntu.com. You may
also use:
apt-cache search <keyword>

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 52/520
 Managing software packages (2)

▶ To install a given package:

sudo apt install <package>
▶ To remove a given package:
sudo apt remove <package>
▶ To install all available package updates:
sudo apt dist-upgrade
▶ Get information about a package:
apt-cache show <package>
▶ Graphical interfaces
▶ Synaptic for GNOME
▶ KPackageKit for KDE

Further details on package management:

https://www.debian.org/doc/manuals/debian-reference/ch02.en.html
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 53/520
 Host vs. target

▶ When doing embedded development, there is always a split between

▶ The host, the development workstation, which is typically a powerful PC
▶ The target, which is the embedded system under development
▶ They are connected by various means: almost always a serial line for debugging
purposes, frequently an Ethernet connection, sometimes a JTAG interface for
low-level debugging

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 54/520
 Serial line communication program

▶ An essential tool for embedded development is a serial line communication

program, like HyperTerminal in Windows.
▶ There are multiple options available in Linux: Minicom, Picocom, Gtkterm, Putty,
screen, etc.
▶ In this training session, we recommend using the simplest of them: picocom
▶ Installation with sudo apt install picocom
▶ Run with picocom -b BAUD_RATE /dev/SERIAL_DEVICE
▶ Exit with Control-A Control-X
▶ SERIAL_DEVICE is typically
▶ ttyUSBx for USB to serial converters
▶ ttySx for real serial ports

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 55/520
 Command line tips

▶ Using the command line is mandatory for many operations needed for embedded
Linux development
▶ It is a very powerful way of interacting with the system, with which you can save
a lot of time.
▶ Some useful tips
▶ You can use several tabs in the Gnome Terminal
▶ Remember that you can use relative paths (for example: ../../linux) in addition
to absolute paths (for example: /home/user)
▶ In a shell, hit [Control] [r], then a keyword, will search through the command
history. Hit [Control] [r] again to search backwards in the history
▶ You can copy/paste paths directly from the file manager to the terminal by
drag-and-drop.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 56/520
 Practical lab - Training Setup

Prepare your lab environment

▶ Download and extract the lab archive

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 57/520
 Cross-compiling toolchains

Cross-compiling
toolchains

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 58/520
 Cross-compiling toolchains

Definition and Components

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 59/520
 Toolchain definition (1)

▶ The usual development tools available on a GNU/Linux workstation is a native

toolchain
▶ This toolchain runs on your workstation and generates code for your workstation,
usually x86
▶ For embedded system development, it is usually impossible or not interesting to
use a native toolchain
▶ The target is too restricted in terms of storage and/or memory
▶ The target is very slow compared to your workstation
▶ You may not want to install all development tools on your target.
▶ Therefore, cross-compiling toolchains are generally used. They run on your
workstation but generate code for your target.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 60/520
 Toolchain definition (2)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 61/520
 Machines in build procedures

▶ Three machines must be distinguished when discussing toolchain creation

▶ The build machine, where the toolchain is built.
▶ The host machine, where the toolchain will be executed.
▶ The target machine, where the binaries created by the toolchain are executed.
▶ Four common build types are possible for toolchains

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 62/520
 Different toolchain build procedures

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 63/520
 Components

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 64/520
 Binutils

▶ Binutils is a set of tools to generate and manipulate binaries for a given CPU
architecture
▶ as, the assembler, that generates binary code from assembler source code
▶ ld, the linker
▶ ar, ranlib, to generate .a archives, used for libraries
▶ objdump, readelf, size, nm, strings, to inspect binaries. Very useful analysis tools!
▶ objcopy, to modify binaries
▶ strip, to strip parts of binaries that are just needed for debugging (reducing their
size).
▶ https://www.gnu.org/software/binutils/
▶ GPL license

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 65/520
 Kernel headers (1)

▶ The C library and compiled programs needs to

interact with the kernel
▶ Available system calls and their numbers
▶ Constant definitions
▶ Data structures, etc.
▶ Therefore, compiling the C library requires
kernel headers, and many applications also
require them.
▶ Available in <linux/...> and <asm/...> and
a few other directories corresponding to the
ones visible in include/ in the kernel sources

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 66/520
 Kernel headers (2)

▶ System call numbers, in <asm/unistd.h>

#define __NR_exit 1
#define __NR_fork 2
#define __NR_read 3
▶ Constant definitions, here in <asm-generic/fcntl.h>, included from
<asm/fcntl.h>, included from <linux/fcntl.h>
#define O_RDWR 00000002
▶ Data structures, here in <asm/stat.h>
struct stat {
unsigned long st_dev;
unsigned long st_ino;
[...]
};

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 67/520
 Kernel headers (3)

▶ The kernel to user space ABI is backward compatible

▶ ABI = Application Binary Interface
We’re talking about binary compatibility
▶ Binaries generated with a toolchain using kernel headers older than the running
kernel will work without problem, but won’t be able to use the new system calls,
data structures, etc.
▶ Binaries generated with a toolchain using kernel headers newer than the running
kernel might work on if they don’t use the recent features, otherwise they will break
▶ Using the latest kernel headers is not necessary, unless access to the new kernel
features is needed
▶ The kernel headers are extracted from the kernel sources using the
headers_install kernel Makefile target.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 68/520
 C/C++ compiler

▶ GCC: GNU Compiler Collection, the famous free software

compiler
▶ https://gcc.gnu.org/
▶ Can compile C, C++, Ada, Fortran, Java, Objective-C,
Objective-C++, Go, etc. Can generate code for a large number
of CPU architectures, including ARM, AVR, Blackfin, CRIS,
FRV, M32, MIPS, MN10300, PowerPC, SH, v850, x86, x86_64,
IA64, Xtensa, etc.
▶ Available under the GPL license, libraries under the GPL with
linking exception.
▶ Alternative: Clang / LLVM compiler
(https://clang.llvm.org/) getting increasingly popular and
able to compile most programs (license: MIT/BSD type)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 69/520
 C library

▶ The C library is an essential component of a

Linux system
by Shmuel Csaba Otto Traian; GNU FDL 1.3 & CC-BY-SA 3.0; created 2014-02-27, last updated 2014-03-25

Application
▶ Interface between the applications and the POSIX-compatible
n
tio
nc lls

kernel fu
ca
BusyBox

▶ Provides the well-known standard C API to

Linux
memory
et al.
manager

m calls
Virtual
IPC syste
file

ease application development

manager
system Linux
process
scheduler
syste
m calls

▶ Several C libraries are available: glibc, uClibc, I/O

interface
Network
interface

fu

musl, klibc, newlib...

nc
ca tion
lls

Linux-specific

▶ The choice of the C library must be made at Application

cross-compiling toolchain generation time, as

the GCC compiler is compiled against a specific Source: Wikipedia (https://bit.ly/2zrGve2)

C library.
Comparing libcs by feature: https://www.etalabs.net/compare_libcs.html

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 70/520
 Cross-compiling toolchains

C Libraries

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 71/520
 glibc

▶ License: LGPL
▶ C library from the GNU project
▶ Designed for performance, standards compliance and
portability
▶ Found on all GNU / Linux host systems
▶ Of course, actively maintained
▶ By default, quite big for small embedded systems. On
armv7hf, version 2.23: libc: 1.5 MB, libm: 492 KB,
source: https://toolchains.bootlin.com
▶ But some features not needed in embedded systems can
be configured out (merged from the old eglibc project).
▶ https://www.gnu.org/software/libc/
Image: https://bit.ly/2EzHl6m

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 72/520
 uClibc-ng

▶ https://uclibc-ng.org/
▶ A continuation of the old uClibc project, license: LGPL
▶ Lightweight C library for small embedded systems
▶ High configurability: many features can be enabled or disabled through a
menuconfig interface.
▶ Supports most embedded architectures, including MMU-less ones (ARM Cortex-M,
Blackfin, etc.). The only library supporting ARM noMMU.
▶ No guaranteed binary compatibility. May need to recompile applications when the
library configuration changes.
▶ Some glibc features may not be implemented yet (real-time, floating-point
operations...)
▶ Focus on size rather than performance
▶ Size on armv7hf, version 1.0.24: libc: 652 KB, source:
https://toolchains.bootlin.com
▶ Actively supported, but Yocto Project stopped supporting it.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 73/520
 musl C library

https://www.musl-libc.org/
▶ A lightweight, fast and simple library for embedded systems
▶ Created while uClibc’s development was stalled
▶ In particular, great at making small static executables
▶ More permissive license (MIT), making it easier to release static
executables. We will talk about the requirements of the LGPL license
(glibc, uClibc) later.
▶ Supported by build systems such as Buildroot and Yocto Project.
▶ Used by the Alpine Linux distribution
(https://www.alpinelinux.org/), fitting in about 130 MB of
storage.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 74/520
 glibc vs uclibc-ng vs musl - small static executables

Let’s compile and strip a hello.c program statically and compare the size
▶ With gcc 6.3, armel, musl 1.1.16:
7300 bytes
▶ With gcc 6.3, armel, uclibc-ng 1.0.22 :
67204 bytes.
▶ With gcc 6.2, armel, glibc:
492792 bytes

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 75/520
 glibc vs uclibc vs musl (static)

Let’s compile and strip BusyBox 1.26.2 statically and compare the size
▶ With gcc 6.3, armel, musl 1.1.16:
183348 bytes
▶ With gcc 6.3, armel, uclibc-ng 1.0.22 :
210620 bytes.
▶ With gcc 6.2, armel, glibc:
755088 bytes
Notes:
▶ BusyBox is automatically compiled with -Os and stripped.
▶ Compiling with shared libraries will mostly eliminate size differences

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 76/520
 Other smaller C libraries

▶ Several other smaller C libraries have been developed, but none of them have the
goal of allowing the compilation of large existing applications
▶ They can run only relatively simple programs, typically to make very small static
executables and run in very small root filesystems.
▶ Choices:
▶ Newlib, https://sourceware.org/newlib/
▶ Klibc, https://kernel.org/pub/linux/libs/klibc/, designed for use in an
initramfs or initrd at boot time.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 77/520
 Advise for choosing the C library

▶ Advice to start developing and debugging your applications with glibc, which is
the most standard solution.
▶ Then, when everything works, if you have size constraints, try to compile your app
and then the entire filesystem with uClibc or musl.
▶ If you run into trouble, it could be because of missing features in the C library.
▶ In case you wish to make static executables, musl will be an easier choice.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 78/520
 Cross-compiling toolchains

Toolchain Options

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 79/520
 ABI

▶ When building a toolchain, the ABI used to generate binaries needs to be defined
▶ ABI, for Application Binary Interface, defines the calling conventions (how
function arguments are passed, how the return value is passed, how system calls
are made) and the organization of structures (alignment, etc.)
▶ All binaries in a system are typically compiled with the same ABI, and the kernel
must understand this ABI.
▶ On ARM, two main ABIs: OABI and EABI
▶ Nowadays everybody uses EABI
▶ On MIPS, several ABIs: o32, o64, n32, n64
▶ https://en.wikipedia.org/wiki/Application_Binary_Interface

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 80/520
 Floating point support

▶ Some processors have a floating point unit, some others do not.

▶ For example, many ARMv4 and ARMv5 CPUs do not have a floating point unit.
Since ARMv7, a VFP unit is mandatory.
▶ For processors having a floating point unit, the toolchain should generate hard
float code, in order to use the floating point instructions directly
▶ For processors without a floating point unit, two solutions
▶ Generate hard float code and rely on the kernel to emulate the floating point
instructions. This is very slow.
▶ Generate soft float code, so that instead of generating floating point instructions,
calls to a user space library are generated
▶ Decision taken at toolchain configuration time
▶ Also possible to configure which floating point unit should be used

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 81/520
 CPU optimization flags

▶ A set of cross-compiling tools is specific to a CPU architecture (ARM, x86, MIPS,

PowerPC)
▶ However, gcc offers further options:
▶ -march allows to select a specific target instruction set
▶ -mtune allows to optimize code for a specific CPU
▶ For example: -march=armv7 -mtune=cortex-a8
▶ -mcpu=cortex-a8 can be used instead to allow gcc to infer the target instruction set
(-march=armv7) and cpu optimizations (-mtune=cortex-a8)
▶ https://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html
▶ At the toolchain compilation time, values can be chosen. They are used:
▶ As the default values for the cross-compiling tools, when no other -march, -mtune,
-mcpu options are passed
▶ To compile the C library
▶ Even if the C library has been compiled for armv5t, it doesn’t prevent from
compiling other programs for armv7
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 82/520
 Cross-compiling toolchains

Obtaining a Toolchain

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 83/520
 Building a toolchain manually

Building a cross-compiling toolchain by yourself is a difficult and painful task! Can

take days or weeks!
▶ Lots of details to learn: many components to build, complicated configuration
▶ Lots of decisions to make (such as C library version, ABI, floating point
mechanisms, component versions)
▶ Need kernel headers and C library sources
▶ Need to be familiar with current gcc issues and patches on your platform
▶ Useful to be familiar with building and configuring tools
▶ See the Crosstool-NG docs/ directory for details on how toolchains are built.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 84/520
 Get a pre-compiled toolchain

▶ Solution that many people choose

▶ Advantage: it is the simplest and most convenient solution
▶ Drawback: you can’t fine tune the toolchain to your needs
▶ Make sure the toolchain you find meets your requirements: CPU, endianness, C
library, component versions, ABI, soft float or hard float, etc.
▶ Possible choices
▶ Toolchains packaged by your distribution
Ubuntu example:
sudo apt install gcc-arm-linux-gnueabihf
▶ Bootlin’s toolchains (for most architectures): https://toolchains.bootlin.com
▶ Toolchain provided by your hardware vendor.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 85/520
 Toolchain building utilities

Another solution is to use utilities that automate the process of building the
toolchain
▶ Same advantage as the pre-compiled toolchains: you don’t need to mess up with
all the details of the build process
▶ But also offers more flexibility in terms of toolchain configuration, component
version selection, etc.
▶ They also usually contain several patches that fix known issues with the different
components on some architectures
▶ Multiple tools with identical principle: shell scripts or Makefile that automatically
fetch, extract, configure, compile and install the different components

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 86/520
 Toolchain building utilities (2)

Crosstool-ng
▶ Rewrite of the older Crosstool, with a
menuconfig-like configuration system
▶ Feature-full: supports uClibc, glibc, musl,
bionic (Android’s C library), hard and soft
float, many architectures
▶ Actively maintained
▶ https://crosstool-ng.github.io/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 87/520
 Toolchain building utilities (3)

Many root filesystem build systems also allow the construction of a cross-compiling
toolchain
▶ Buildroot
▶ Makefile-based. Can build glibc, uClibc and musl based toolchains, for a wide range
of architectures.
▶ https://buildroot.org
▶ PTXdist
▶ Makefile-based, maintained mainly by Pengutronix. It only supports uClibc and glibc
(version 2020.05 status)
▶ https://www.ptxdist.org/
▶ OpenEmbedded / Yocto Project
▶ A featureful, but more complicated build system
▶ http://www.openembedded.org/
▶ https://www.yoctoproject.org/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 88/520
 Crosstool-NG: installation and usage
▶ Installation of Crosstool-NG can be done system-wide, or just locally in the source
directory. For local installation:
./configure --enable-local
make
make install
▶ Some sample configurations for various architectures are available in samples, they
can be listed using
./ct-ng list-samples
▶ To load a sample configuration
./ct-ng <sample-name>
▶ To adjust the configuration
./ct-ng menuconfig
▶ To build the toolchain
./ct-ng build
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 89/520
 Toolchain contents

▶ The cross compilation tool binaries, in bin/

▶ This directory should be added to your PATH to ease usage of the toolchain
▶ One or several sysroot, each containing
▶ The C library and related libraries, compiled for the target
▶ The C library headers and kernel headers
▶ There is one sysroot for each variant: toolchains can be multilib if they have
several copies of the C library for different configurations (for example: ARMv4T,
ARMv5T, etc.)
▶ Old CodeSourcery ARM toolchains were multilib, the sysroots in:
arm-none-linux-gnueabi/libc/, arm-none-linux-gnueabi/libc/armv4t/,
arm-none-linux-gnueabi/libc/thumb2
▶ Crosstool-NG toolchains can be multilib too (still experimental), otherwise the
sysroot is in arm-unknown-linux-uclibcgnueabi/sysroot

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 90/520
 Practical lab - Using Crosstool-NG

Time to build your toolchain

▶ Configure Crosstool-NG
▶ Run it to build your own cross-compiling
toolchain

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 91/520
 Bootloaders

Bootloaders

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 92/520
 Bootloaders

Boot Sequence

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 93/520
 Bootloaders

▶ The bootloader is a piece of code responsible for

▶ Basic hardware initialization
▶ Loading of an application binary, usually an operating system kernel, from flash
storage, from the network, or from another type of non-volatile storage.
▶ Possibly decompression of the application binary
▶ Execution of the application
▶ Besides these basic functions, most bootloaders provide a shell with various
commands implementing different operations.
▶ Loading of data from storage or network, memory inspection, hardware diagnostics
and testing, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 94/520
 Bootloaders on BIOS-based x86 (1)

▶ The x86 processors are typically bundled on a board with a

non-volatile memory containing a program, the BIOS.
▶ On old BIOS-based x86 platforms: the BIOS is responsible for
basic hardware initialization and loading of a very small piece of
code from non-volatile storage.
▶ This piece of code is typically a 1st stage bootloader, which will
load the full bootloader itself.
▶ It typically understands filesystem formats so that the kernel file
can be loaded directly from a normal filesystem.
▶ This sequence is different for modern EFI-based systems.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 95/520
 Bootloaders on x86 (2)

▶ GRUB, Grand Unified Bootloader, the most powerful one.

https://www.gnu.org/software/grub/
▶ Can read many filesystem formats to load the kernel image and the configuration,
provides a powerful shell with various commands, can load kernel images over the
network, etc.
▶ See our dedicated presentation for details:
https://bootlin.com/doc/legacy/grub/
▶ Syslinux, for network and removable media booting (USB key, CD-ROM)
https://kernel.org/pub/linux/utils/boot/syslinux/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 96/520
 Booting on embedded CPUs: case 1

▶ When powered, the CPU starts executing code at a fixed address

▶ There is no other booting mechanism provided by the CPU
▶ The hardware design must ensure that a NOR flash chip is wired
so that it is accessible at the address at which the CPU starts
executing instructions
▶ The first stage bootloader must be programmed at this address
in the NOR
▶ NOR is mandatory, because it allows random access, which
NAND doesn’t allow
▶ Not very common anymore (unpractical, and requires NOR
flash)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 97/520
 Booting on embedded CPUs: case 2

▶ The CPU has an integrated boot code in ROM

▶ BootROM on AT91 CPUs, “ROM code” on OMAP, etc.
▶ Exact details are CPU-dependent
▶ This boot code is able to load a first stage bootloader from a storage device into
an internal SRAM (DRAM not initialized yet)
▶ Storage device can typically be: MMC, NAND, SPI flash, UART (transmitting data
over the serial line), etc.
▶ The first stage bootloader is
▶ Limited in size due to hardware constraints (SRAM size)
▶ Provided either by the CPU vendor or through community projects
▶ This first stage bootloader must initialize DRAM and other hardware devices and
load a second stage bootloader into RAM

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 98/520
 Booting on Microchip ARM SAMA5D3

▶ RomBoot: tries to find a valid bootstrap image from various

storage sources, and load it into SRAM (DRAM not initialized
yet). Size limited to 64 KB. No user interaction possible in
standard boot mode.
▶ AT91Bootstrap: runs from SRAM. Initializes the DRAM, the
NAND or SPI controller, and loads the secondary bootloader into
RAM and starts it. No user interaction possible.
▶ U-Boot: runs from RAM. Initializes some other hardware devices
(network, USB, etc.). Loads the kernel image from storage or
network to RAM and starts it. Shell with commands provided.
▶ Linux Kernel: runs from RAM. Takes over the system completely
(the bootloader no longer exists).
Note: same process on other Microchip AT91 SoCs, but the SRAM size
is smaller on the older ones.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 99/520
 Booting on ARM TI OMAP2+ / AM33xx

▶ ROM Code: tries to find a valid bootstrap image from various

storage sources, and load it into SRAM or RAM (RAM can be
initialized by ROM code through a configuration header). Size
limited to <64 KB. No user interaction possible.
▶ X-Loader or U-Boot SPL: runs from SRAM. Initializes the
DRAM, the NAND or MMC controller, and loads the secondary
bootloader into RAM and starts it. No user interaction possible.
File called MLO.
▶ U-Boot: runs from RAM. Initializes some other hardware devices
(network, USB, etc.). Loads the kernel image from storage or
network to RAM and starts it. Shell with commands provided.
File called u-boot.bin or u-boot.img.
▶ Linux Kernel: runs from RAM. Takes over the system completely
(bootloaders no longer exists).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 100/520
 Booting on Marvell SoCs

▶ ROM Code: tries to find a valid bootstrap image from various

storage sources, and load it into RAM. The RAM configuration is
described in a CPU-specific header, prepended to the bootloader
image.
▶ U-Boot: runs from RAM. Initializes some other hardware devices
(network, USB, etc.). Loads the kernel image from storage or
network to RAM and starts it. Shell with commands provided.
File called u-boot.kwb.
▶ Linux Kernel: runs from RAM. Takes over the system completely
(bootloaders no longer exists).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 101/520
 Generic bootloaders for embedded CPUs

▶ We will focus on the generic part, the main bootloader, offering the most
important features.
▶ There are several open-source generic bootloaders.
Here are the most popular ones:
▶ U-Boot, the universal bootloader by Denx
The most used on ARM, also used on PPC, MIPS, x86, m68k, RiscV, etc. The
de-facto standard nowadays. We will study it in detail.
https://www.denx.de/wiki/U-Boot
▶ Barebox, an architecture-neutral bootloader, written as a successor of U-Boot. It
doesn’t have as much hardware support as U-Boot yet. U-Boot has improved quite a
lot thanks to this competitor.
https://www.barebox.org

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 102/520
 Bootloaders

The U-boot bootloader

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 103/520
 U-Boot

U-Boot is a typical free software project

▶ License: GPLv2 (same as Linux)
▶ Freely available at https://www.denx.de/wiki/U-Boot
▶ Documentation available at https://www.denx.de/wiki/U-Boot/Documentation
▶ The latest development source code is available in a Git repository:
https://gitlab.denx.de/u-boot/u-boot
▶ Development and discussions happen around an open mailing-list
https://lists.denx.de/pipermail/u-boot/
▶ Follows a fixed-interval release schedule. Every two months, a new version is
released. Versions are named YYYY.MM.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 104/520
 U-Boot configuration

▶ Get the source code from the website, and uncompress it

▶ The configs/ directory contains one configuration file for each supported board
▶ It defines the CPU type, the peripherals and their configuration, the memory
mapping, the U-Boot features that should be compiled in, etc.
▶ Note: U-Boot is migrating from board configuration defined in C header files
(include/configs/) to defconfig like in the Linux kernel (configs/)
▶ Not all boards have been converted to the new configuration system.
▶ Older U-Boot releases provided by hardware vendors may not yet use this new
configuration system.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 105/520
 U-Boot configuration file

CHIP_defconfig
CONFIG_ARM=y
CONFIG_ARCH_SUNXI=y
CONFIG_MACH_SUN5I=y
CONFIG_DRAM_TIMINGS_DDR3_800E_1066G_1333J=y
# CONFIG_MMC is not set
CONFIG_USB0_VBUS_PIN="PB10"
CONFIG_VIDEO_COMPOSITE=y
CONFIG_DEFAULT_DEVICE_TREE="sun5i-r8-chip"
CONFIG_SPL=y
CONFIG_SYS_EXTRA_OPTIONS="CONS_INDEX=2"
# CONFIG_CMD_IMLS is not set
CONFIG_CMD_DFU=y
CONFIG_CMD_USB_MASS_STORAGE=y
CONFIG_AXP_ALDO3_VOLT=3300
CONFIG_AXP_ALDO4_VOLT=3300
CONFIG_USB_MUSB_GADGET=y
CONFIG_USB_GADGET=y
CONFIG_USB_GADGET_DOWNLOAD=y
CONFIG_G_DNL_MANUFACTURER="Allwinner Technology"
CONFIG_G_DNL_VENDOR_NUM=0x1f3a
CONFIG_G_DNL_PRODUCT_NUM=0x1010
CONFIG_USB_EHCI_HCD=y

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 106/520
 Configuring and compiling U-Boot

▶ U-Boot must be configured before being compiled

▶ make BOARDNAME_defconfig
▶ Where BOARDNAME is the name of a configuration, as visible in the configs/
directory.
▶ You can then run make menuconfig to further customize U-Boot’s configuration!
▶ Make sure that the cross-compiler is available in PATH
▶ Compile U-Boot, by specifying the cross-compiler prefix.
Example, if your cross-compiler executable is arm-linux-gcc:
make CROSS_COMPILE=arm-linux-
▶ The main result is a u-boot.bin file, which is the U-Boot image. Depending on
your specific platform, there may be other specialized images: u-boot.img,
u-boot.kwb, MLO, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 107/520
 Installing U-Boot

U-Boot must usually be installed in flash memory to be executed by the hardware.

Depending on the hardware, the installation of U-Boot is done in a different way:
▶ The CPU provides some kind of specific boot monitor with which you can
communicate through serial port or USB using a specific protocol
▶ The CPU boots first on removable media (MMC) before booting from fixed media
(NAND). In this case, boot from MMC to reflash a new version
▶ U-Boot is already installed, and can be used to flash a new version of U-Boot.
However, be careful: if the new version of U-Boot doesn’t work, the board is
unusable
▶ The board provides a JTAG interface, which allows to write to the flash memory
remotely, without any system running on the board. It also allows to rescue a
board if the bootloader doesn’t work.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 108/520
 U-boot prompt
▶ Connect the target to the host through a serial console.
▶ Power-up the board. On the serial console, you will see something like:
U-Boot 2020.04 (May 26 2020 - 16:05:43 +0200)

CPU: SAMA5D36
Crystal frequency: 12 MHz
CPU clock : 528 MHz
Master clock : 132 MHz
DRAM: 256 MiB
NAND: 256 MiB
MMC: Atmel mci: 0, Atmel mci: 1
Loading Environment from NAND... OK
In: serial@ffffee00
Out: serial@ffffee00
Err: serial@ffffee00
Net: eth0: ethernet@f0028000
Error: ethernet@f802c000 address not set.

Hit any key to stop autoboot: 0

=>

▶ The U-Boot shell offers a set of commands. We will study the most important
ones, see the documentation for a complete reference or the help command.
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 109/520
 Information commands (1)

Flash information (NOR and SPI flash)

=> flinfo
DataFlash:AT45DB021
Nb pages: 1024
Page Size: 264
Size= 270336 bytes
Logical address: 0xC0000000
Area 0: C0000000 to C0001FFF (RO) Bootstrap
Area 1: C0002000 to C0003FFF Environment
Area 2: C0004000 to C0041FFF (RO) U-Boot

NAND flash information

=> nand info

Device 0: nand0, sector size 128 KiB

Page size 2048 b
OOB size 64 b
Erase size 131072 b
subpagesize 2048 b
options 0x40004200
bbt options 0x00008000

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 110/520
 Information commands (2)

Version details
=> version
U-Boot 2020.04 (May 26 2020 - 16:05:43 +0200)

arm-linux-gcc (crosstool-NG 1.24.0.105_5659366) 9.2.0

GNU ld (crosstool-NG 1.24.0.105_5659366) 2.34

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 111/520
 Important commands (1)

▶ The exact set of commands depends on the U-Boot configuration

▶ help and help command
▶ fatload, loads a file from a FAT filesystem to RAM
▶ Example: fatload usb 0:1 0x21000000 zImage
▶ And also fatls and fatinfo
▶ ext2load, loads a file from an ext2 filesystem to RAM
▶ And also ext2ls to list files, ext2info for information
▶ tftp, loads a file from the network to RAM
▶ ping, to test the network
▶ boot, runs the default boot command, stored in bootcmd
▶ bootz <address>, starts a kernel image loaded at the given address in RAM

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 112/520
 Important commands (2)

▶ loadb, loads, loady, load a file from the serial line to RAM
▶ usb, to initialize and control the USB subsystem, mainly used for USB storage
devices such as USB keys
▶ mmc, to initialize and control the MMC subsystem, used for SD and microSD cards
▶ nand, to erase, read and write contents to NAND flash
▶ erase, protect, cp, to erase, modify protection and write to NOR flash
▶ md, displays memory contents. Can be useful to check the contents loaded in
memory, or to look at hardware registers.
▶ mm, modifies memory contents. Can be useful to modify directly hardware
registers, for testing purposes.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 113/520
 Environment variables: principle

▶ U-Boot can be configured through environment variables

▶ Some specific environment variables affect the behavior of the different commands
▶ Custom environment variables can be added, and used in scripts
▶ Environment variables are loaded from flash to RAM at U-Boot startup, can be
modified and saved back to flash for persistence
▶ There is a dedicated location in flash (or in MMC storage) to store the U-Boot
environment, defined in the board configuration file

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 114/520
 Environment variables commands (2)

Commands to manipulate environment variables:

▶ printenv
Shows all variables
▶ printenv <variable-name>
Shows the value of a variable
▶ setenv <variable-name> <variable-value>
Changes the value of a variable, only in RAM
▶ editenv <variable-name>
Edits the value of a variable, only in RAM
▶ saveenv
Saves the current state of the environment to flash

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 115/520
 Environment variables commands - Example

u-boot # printenv
baudrate=19200
ethaddr=00:40:95:36:35:33
netmask=255.255.255.0
ipaddr=10.0.0.11
serverip=10.0.0.1
stdin=serial
stdout=serial
stderr=serial
u-boot # printenv serverip
serverip=10.0.0.1
u-boot # setenv serverip 10.0.0.100
u-boot # saveenv

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 116/520
 Important U-Boot env variables

▶ bootcmd, specifies the commands that U-Boot will automatically execute at boot
time after a configurable delay (bootdelay), if the process is not interrupted
▶ bootargs, contains the arguments passed to the Linux kernel, covered later
▶ serverip, the IP address of the server that U-Boot will contact for network
related commands
▶ ipaddr, the IP address that U-Boot will use
▶ netmask, the network mask to contact the server
▶ ethaddr, the MAC address, can only be set once
▶ autostart, if set to yes, U-Boot automatically starts an image after loading it in
memory (tftp, fatload...)
▶ filesize, the size of the latest copy to memory (from tftp, fatload,
nand read...)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 117/520
 Scripts in environment variables

▶ Environment variables can contain small scripts, to execute several commands and
test the results of commands.
▶ Useful to automate booting or upgrade processes
▶ Several commands can be chained using the ; operator
▶ Tests can be done using if command ; then ... ; else ... ; fi
▶ Scripts are executed using run <variable-name>
▶ You can reference other variables using ${variable-name}
▶ Example
▶ setenv mmc-boot 'if fatload mmc 0 80000000 boot.ini; then source; else
if fatload mmc 0 80000000 zImage; then run mmc-do-boot; fi; fi'

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 118/520
 Transferring files to the target

▶ U-Boot is mostly used to load and boot a kernel image, but it also allows to
change the kernel image and the root filesystem stored in flash.
▶ Files must be exchanged between the target and the development workstation.
This is possible:
▶ Through the network if the target has an Ethernet connection, and U-Boot contains
a driver for the Ethernet chip. This is the fastest and most efficient solution.
▶ Through a USB key, if U-Boot supports the USB controller of your platform
▶ Through a SD or microSD card, if U-Boot supports the MMC controller of your
platform
▶ Through the serial port (loadb, loadx or loady command)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 119/520
 TFTP

▶ Network transfer from the development workstation to U-Boot on the target takes
place through TFTP
▶ Trivial File Transfer Protocol
▶ Somewhat similar to FTP, but without authentication and over UDP
▶ A TFTP server is needed on the development workstation
▶ sudo apt install tftpd-hpa
▶ All files in /var/lib/tftpboot (in Ubuntu 18.04) or in /srv/tftp (in Ubuntu
20.04) are then visible through TFTP
▶ A TFTP client is available in the tftp-hpa package, for testing
▶ A TFTP client is integrated into U-Boot
▶ Configure the ipaddr and serverip environment variables
▶ Use tftp <address> <filename> to load a file

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 120/520
 Practical lab - U-Boot

Time to start the practical lab!

▶ Communicate with the board using a serial
console
▶ Configure, build and install the first stage
bootloader
▶ Configure, build and install U-Boot
▶ Learn U-Boot commands
▶ Set up TFTP communication with the board

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 121/520
 Linux kernel introduction

Linux kernel
introduction

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 122/520
 Linux kernel introduction

Linux features

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 123/520
 History

▶ The Linux kernel is one component of a system, which also

requires libraries and applications to provide features to end
users.
▶ The Linux kernel was created as a hobby in 1991 by a Finnish
student, Linus Torvalds.
▶ Linux quickly started to be used as the kernel for free software
operating systems
▶ Linus Torvalds has been able to create a large and dynamic
developer and user community around Linux. Linus Torvalds in 2014
▶ Nowadays, more than one thousand people contribute to each Image credits (Wikipedia):
https://bit.ly/2UIa1TD
kernel release, individuals or companies big and small.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 124/520
 Linux kernel key features

▶ Portability and hardware support. ▶ Security. It can’t hide its flaws. Its
Runs on most architectures. code is reviewed by many experts.
▶ Scalability. Can run on super ▶ Stability and reliability.
computers as well as on tiny devices ▶ Modularity. Can include only what a
(4 MB of RAM is enough). system needs even at run time.
▶ Compliance to standards and ▶ Easy to program. You can learn from
interoperability. existing code. Many useful resources
▶ Exhaustive networking support. on the net.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 125/520
 Linux kernel in the system

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 126/520
 Linux kernel main roles

▶ Manage all the hardware resources: CPU, memory, I/O.

▶ Provide a set of portable, architecture and hardware independent APIs to
allow user space applications and libraries to use the hardware resources.
▶ Handle concurrent accesses and usage of hardware resources from different
applications.
▶ Example: a single network interface is used by multiple user space applications
through various network connections. The kernel is responsible to “multiplex” the
hardware resource.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 127/520
 System calls

▶ The main interface between the kernel and user space is

the set of system calls
▶ About 400 system calls that provide the main kernel
services
▶ File and device operations, networking operations,
inter-process communication, process management,
memory mapping, timers, threads, synchronization
primitives, etc.
▶ This interface is stable over time: only new system calls
can be added by the kernel developers
▶ This system call interface is wrapped by the C library,
Image credits (Wikipedia):
and user space applications usually never make a system https://bit.ly/2U2rdGB
call directly but rather use the corresponding C library
function
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 128/520
 Pseudo filesystems

▶ Linux makes system and kernel information available in user space through
pseudo filesystems, sometimes also called virtual filesystems
▶ Pseudo filesystems allow applications to see directories and files that do not exist
on any real storage: they are created and updated on the fly by the kernel
▶ The two most important pseudo filesystems are
▶ proc, usually mounted on /proc:
Operating system related information (processes, memory management
parameters...)
▶ sysfs, usually mounted on /sys:
Representation of the system as a set of devices and buses. Information about these
devices.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 129/520
 Inside the Linux kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 130/520
 Linux license

▶ The whole Linux sources are Free Software released under the GNU General
Public License version 2 (GPL v2).
▶ For the Linux kernel, this basically implies that:
▶ When you receive or buy a device with Linux on it, you should receive the Linux
sources, with the right to study, modify and redistribute them.
▶ When you produce Linux based devices, you must release the sources to the
recipient, with the same rights, with no restriction.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 131/520
 Supported hardware architectures

See the arch/ directory in the kernel sources

▶ Minimum: 32 bit processors, with or without MMU, and gcc support
▶ 32 bit architectures (arch/ subdirectories)
Examples: arm, arc, c6x, m68k, microblaze...
▶ 64 bit architectures:
Examples: alpha, arm64, ia64...
▶ 32/64 bit architectures
Examples: mips, powerpc, riscv, sh, sparc, x86...
▶ Note that unmaintained architectures can also be removed when they have
compiling issues and nobody fixes them.
▶ Find details in kernel sources: arch/<arch>/Kconfig, arch/<arch>/README, or
Documentation/<arch>/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 132/520
 Linux kernel introduction

Linux versioning scheme and development

process

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 133/520
 Linux versioning scheme

▶ Until 2003, there was a new stable release branch of Linux every 2 or 3 years (2.0,
2.2, 2.4). New development branches took 2-3 years to become stable (too slow!).
▶ Since 2003, there is a new stable release of Linux about every 10 weeks:
▶ Versions 2.6 (Dec. 2003) to 2.6.39 (May 2011)
▶ Versions 3.0 (Jul. 2011) to 3.19 (Feb. 2015)
▶ Versions 4.0 (Apr. 2015) to 4.20 (Dec. 2018)
▶ Version 5.0 was released in Mar. 2019.
▶ Features are added to the kernel in a progressive way. Since 2003, kernel
developers have managed to do so without having to introduce a massively
incompatible development branch.
▶ For each release, there are bugfix and security updates: 5.0.1, 5.0.2, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 134/520
 Linux development model

Using merge and bug fixing windows

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 135/520
 Need for long term support (1)

▶ Issue: bug and security fixes only released for most recent stable kernel versions.
Only LTS (Long Term Support) releases are supported for up to 6 years.
▶ Example at Google: starting from Android O (2017), all new Android devices will
have to run such an LTS kernel.

Captured on https://kernel.org in Nov.

2019, following the Releases link.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 136/520
 Need for long term support (2)

▶ You could also get long term support from a commercial embedded Linux provider.
▶ ”If you are not using a supported distribution kernel, or a stable / longterm kernel,
you have an insecure kernel” - Greg KH, 2019
Some vulnerabilities are fixed in stable without ever getting a CVE.
▶ The Civil Infrastructure Platform project is an industry / Linux Foundation effort
to support selected LTS versions (starting with 4.4) much longer (> 10 years).
See https://bit.ly/2hy1QYC.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 137/520
 What’s new in each Linux release? (1)

The official list of changes for each Linux release is just a huge list of individual
patches!
commit aa6e52a35d388e730f4df0ec2ec48294590cc459
Author: Thomas Petazzoni <[email protected]>
Date: Wed Jul 13 11:29:17 2011 +0200

at91: at91-ohci: support overcurrent notification

Several USB power switches (AIC1526 or MIC2026) have a digital output

that is used to notify that an overcurrent situation is taking
place. This digital outputs are typically connected to GPIO inputs of
the processor and can be used to be notified of these overcurrent
situations.

Therefore, we add a new overcurrent_pin[] array in the at91_usbh_data

structure so that boards can tell the AT91 OHCI driver which pins are
used for the overcurrent notification, and an overcurrent_supported
boolean to tell the driver whether overcurrent is supported or not.

The code has been largely borrowed from ohci-da8xx.c and

ohci-s3c2410.c.

Signed-off-by: Thomas Petazzoni <[email protected]>

Signed-off-by: Nicolas Ferre <[email protected]>

Very difficult to find out the key changes and to get the global picture out of individual
changes.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 138/520
 What’s new in each Linux release? (2)

Fortunately, there are some useful resources available

▶ https://kernelnewbies.org/LinuxChanges
In depth coverage of the new features in each kernel release
▶ https://lwn.net
Coverage of the features accepted in each merge window
▶ https://www.linux-arm.info
News about Linux on ARM, including kernel changes.
▶ https://linuxfr.org, for French readers. There is also a long summary of new
features (different from the content on LinuxChanges).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 139/520
 Linux kernel introduction

Linux kernel sources

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 140/520
 Location of kernel sources

▶ The official (mainline) versions of the Linux kernel, as released by Linus Torvalds,
are available at https://kernel.org
▶ These versions follow the development model of the kernel
▶ However, they may not contain the latest development from a specific area yet.
Some features in development might not be ready for mainline inclusion yet.
▶ Many chip vendors supply their own kernel sources
▶ Focusing on hardware support first
▶ Can have a very important delta with mainline Linux
▶ Useful only when mainline hasn’t caught up yet.
▶ Many kernel sub-communities maintain their own kernel, with usually newer but
less stable features
▶ Architecture communities (ARM, MIPS, PowerPC, etc.), device drivers communities
(I2C, SPI, USB, PCI, network, etc.), other communities (real-time, etc.)
▶ No official releases, only meant for sharing work and contributing to the mainline
version.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 141/520
 Getting Linux sources

▶ The kernel sources are available from https://kernel.org/pub/linux/kernel

as full tarballs (complete kernel sources) and patches (differences between two
kernel versions).
▶ However, more and more people use the git version control system. Absolutely
needed for kernel development!
▶ Fetch the entire kernel sources and history
git clone git:
//git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
▶ Create a branch that starts at a specific stable version
git checkout -b <name-of-branch> v5.6
▶ Web interface available at
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/
▶ Read more about Git at https://git-scm.com/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 142/520
 Working with git: SSD storage needed for serious work

For all work with git, but especially on big projects

such as the Linux kernel...
▶ Having a fast disk will dramatically speed up
most git operations.
▶ Ask your boss to order an SSD disk for your
laptop. It will make you more productive.
▶ If you are in a Bootlin public session, you
already have an SSD disk!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 143/520
 Linux kernel size (1)

▶ Linux 5.4 sources:

▶ 66031 files (git ls-files | wc -l)
▶ 27679764 lines (git ls-files | xargs cat | wc -l)
▶ 889221135 bytes (git ls-files | xargs cat | wc -c)
▶ A minimum uncompressed Linux kernel just sizes 1-2 MB
▶ Why are these sources so big?
Because they include thousands of device drivers, many network protocols,
support many architectures and filesystems...
▶ The Linux core (scheduler, memory management...) is pretty small!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 144/520
 Linux kernel size (2)

As of kernel version 5.7 (in percentage of total number of lines).

▶ drivers/: 60.1% ▶ lib/: 0.6%

▶ arch/: 12.9% ▶ mm/: 0.5%
▶ fs/: 4.7% ▶ scripts/: 0.4%
▶ sound/: 4.2% ▶ crypto/: 0.4%
▶ net/: 4.0% ▶ security/: 0.3%
▶ include/: 3.6% ▶ block/: 0.2%
▶ tools/: 3.2% ▶ samples/: 0.1%
▶ Documentation/: 3.2% ▶ virt/: 0.1%
▶ kernel/: 1.3% ▶ ...

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 145/520
 Getting Linux sources
▶ Full tarballs
▶ Contain the complete kernel sources: long to download and uncompress, but must
be done at least once
▶ Example:
https://kernel.org/pub/linux/kernel/v4.x/linux-4.20.13.tar.xz
▶ Extract command:
tar xf linux-4.20.13.tar.xz
▶ Incremental patches between versions
▶ It assumes you already have a base version and you apply the correct patches in the
right order. Quick to download and apply
▶ Examples:
https://kernel.org/pub/linux/kernel/v4.x/patch-4.20.xz
(4.19 to 4.20)
https://kernel.org/pub/linux/kernel/v4.x/patch-4.20.13.xz
(4.20 to 4.20.13)
▶ All previous kernel versions are available in
https://kernel.org/pub/linux/kernel/
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 146/520
 Patch
▶ A patch is the difference between two source trees
▶ Computed with the diff tool, or with more elaborate version control systems
▶ They are very common in the open-source community.
See https://en.wikipedia.org/wiki/Diff
▶ Excerpt from a patch:
diff -Nru a/Makefile b/Makefile
--- a/Makefile 2005-03-04 09:27:15 -08:00
+++ b/Makefile 2005-03-04 09:27:15 -08:00
@@ -1,7 +1,7 @@
VERSION = 2
PATCHLEVEL = 6
SUBLEVEL = 11
-EXTRAVERSION =
+EXTRAVERSION = .1
NAME=Woozy Numbat

# *DOCUMENTATION*
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 147/520
 Contents of a patch
▶ One section per modified file, starting with a header
diff -Nru a/Makefile b/Makefile
--- a/Makefile 2005-03-04 09:27:15 -08:00
+++ b/Makefile 2005-03-04 09:27:15 -08:00
▶ One sub-section (hunk) per modified part of the file, starting with a header with the
starting line number and the number of lines the change hunk applies to
@@ -1,7 +1,7 @@
▶ Three lines of context before the change
VERSION = 2
PATCHLEVEL = 6
SUBLEVEL = 11
▶ The change itself
-EXTRAVERSION =
+EXTRAVERSION = .1
▶ Three lines of context after the change
NAME=Woozy Numbat
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 148/520
 Using the patch command

The patch command:

▶ Takes the patch contents on its standard input
▶ Applies the modifications described by the patch into the current directory
patch usage examples:
▶ patch -p<n> < diff_file
▶ cat diff_file | patch -p<n>
▶ xzcat diff_file.xz | patch -p<n>
▶ zcat diff_file.gz | patch -p<n>
▶ Notes:
▶ n: number of directory levels to skip in the file paths
▶ You can reverse apply a patch with the -R option
▶ You can test a patch with --dry-run option

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 149/520
 Applying a Linux patch
▶ Two types of Linux patches:
▶ Either to be applied to the previous stable version
(from x.<y-1> to x.y)
▶ Or implementing fixes to the current stable version
(from x.y to x.y.z)
▶ Can be downloaded in gzip or xz (much smaller) compressed files.
▶ Always produced for n=1
(that’s what everybody does... do it too!)
▶ Need to run the patch command inside the toplevel kernel source directory
▶ Linux patch command line example:
cd linux-4.19
xzcat ../patch-4.20.xz | patch -p1
xzcat ../patch-4.20.13.xz | patch -p1
cd ..; mv linux-4.19 linux-4.20.13
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 150/520
 Practical lab - Kernel sources

Time to start the practical lab!

▶ Get the Linux kernel sources
▶ Apply patches

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 151/520
 Linux kernel introduction

Building the kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 152/520
 Linux kernel introduction

Kernel configuration

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 153/520
 Kernel configuration

▶ The kernel contains thousands of device drivers, filesystem drivers, network

protocols and other configurable items
▶ Thousands of options are available, that are used to selectively compile parts of
the kernel source code
▶ The kernel configuration is the process of defining the set of options with which
you want your kernel to be compiled
▶ The set of options depends
▶ On the target architecture and on your hardware (for device drivers, etc.)
▶ On the capabilities you would like to give to your kernel (network capabilities,
filesystems, real-time, etc.). Such generic options are available in all architectures.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 154/520
 Specifying the target architecture

First, specify the architecture for the kernel to build

▶ Set it to the name of a directory under arch/:
export ARCH=arm
▶ By default, the kernel build system assumes that the kernel is configured and built
for the host architecture (x86 in our case, native kernel compiling)
▶ The kernel build system will use this setting to:
▶ Use the configuration options for the target architecture.
▶ Compile the kernel with source code and headers for the target architecture.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 155/520
 Kernel configuration and build system

▶ The kernel configuration and build system is based on multiple Makefiles

▶ One only interacts with the main Makefile, present at the top directory of the
kernel source tree
▶ Interaction takes place
▶ using the make tool, which parses the Makefile
▶ through various targets, defining which action should be done (configuration,
compilation, installation, etc.). Run make help to see all available targets.
▶ Example
▶ cd linux-4.14.x/
▶ make <target>

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 156/520
 Kernel configuration details

▶ The configuration is stored in the .config file at the root of kernel sources
▶ Simple text file, CONFIG_PARAM=value (included by the kernel Makefile)
▶ As options have dependencies, typically never edited by hand, but through
graphical or text interfaces:
▶ make xconfig, make gconfig (graphical)
▶ make menuconfig, make nconfig (text)
▶ You can switch from one to another, they all load/save the same .config file, and
show the same set of options

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 157/520
 Initial configuration

Difficult to find which kernel configuration will work with your hardware and root
filesystem. Start with one that works!
▶ Desktop or server case:
▶ Advisable to start with the configuration of your running kernel, usually available in
/boot:
cp /boot/config-`uname -r` .config
▶ Embedded platform case:
▶ Default configuration files are available, usually for each CPU family.
▶ They are stored in arch/<arch>/configs/, and are just minimal .config files (only
settings different from default ones).
▶ Run make help to find if one is available for your platform
▶ To load a default configuration file, just run
make cpu_defconfig
▶ This will overwrite your existing .config file!
Now, you can make configuration changes (make menuconfig...).
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 158/520
 Create your own default configuration

To create your own default configuration file:

▶ make savedefconfig
This creates a minimal configuration (non-default settings)
▶ mv defconfig arch/<arch>/configs/myown_defconfig
This way, you can share a reference configuration inside the kernel sources.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 159/520
 Kernel or module?

▶ The kernel image is a single file, resulting from the linking of all object files that
correspond to features enabled in the configuration
▶ This is the file that gets loaded in memory by the bootloader
▶ All included features are therefore available as soon as the kernel starts, at a time
where no filesystem exists
▶ Some features (device drivers, filesystems, etc.) can however be compiled as
modules
▶ These are plugins that can be loaded/unloaded dynamically to add/remove features
to the kernel
▶ Each module is stored as a separate file in the filesystem, and therefore access
to a filesystem is mandatory to use modules
▶ This is not possible in the early boot procedure of the kernel, because no filesystem
is available

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 160/520
 Kernel option types

There are different types of options, defined in Kconfig files:

▶ bool options, they are either
▶ true (to include the feature in the kernel) or
▶ false (to exclude the feature from the kernel)
▶ tristate options, they are either
▶ true (to include the feature in the kernel image) or
▶ module (to include the feature as a kernel module) or
▶ false (to exclude the feature)
▶ int options, to specify integer values
▶ hex options, to specify hexadecimal values
▶ string options, to specify string values

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 161/520
 Kernel option dependencies
There are dependencies between kernel options
▶ For example, enabling a network driver requires the network stack to be enabled
▶ Two types of dependencies:
▶ depends on dependencies. In this case,
option B that depends on option A is not
visible until option A is enabled
▶ select dependencies. In this case, with
option B depending on option A, when
option A is enabled, option B is
automatically enabled. In particular, such
dependencies are used to declare what
features a hardware architecture supports.
▶ With the Show All Options option, make xconfig allows to see all options, even
the ones that cannot be selected because of missing dependencies. Values for
dependencies are shown.
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 162/520
 make xconfig

make xconfig
▶ The most common graphical interface to configure the kernel.
▶ File browser: easier to load configuration files
▶ Search interface to look for parameters
▶ Required Debian / Ubuntu packages: qt5-default

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 163/520
 make xconfig screenshot

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 164/520
 make xconfig search interface
Looks for a keyword in the parameter name (shortcut: [Ctrl] + [f]).
Allows to set values to found parameters.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 165/520
 Kernel configuration options

Compiled as a module (separate file)

CONFIG_ISO9660_FS=m

Driver options
CONFIG_JOLIET=y
CONFIG_ZISOFS=y

Compiled statically into the kernel

CONFIG_UDF_FS=y

Values in resulting .config file Parameter values as displayed in make xconfig

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 166/520
 Corresponding .config file excerpt
Options are grouped by sections and are prefixed with CONFIG_.
#
# CD-ROM/DVD Filesystems
#
CONFIG_ISO9660_FS=m
CONFIG_JOLIET=y
CONFIG_ZISOFS=y
CONFIG_UDF_FS=y
CONFIG_UDF_NLS=y

#
# DOS/FAT/NT Filesystems
#
# CONFIG_MSDOS_FS is not set
# CONFIG_VFAT_FS is not set
CONFIG_NTFS_FS=m
# CONFIG_NTFS_DEBUG is not set
CONFIG_NTFS_RW=y

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 167/520
 make gconfig

make gconfig
▶ GTK based graphical configuration
interface. Functionality similar to that
of make xconfig.
▶ Just lacking a search functionality.
▶ Required Debian packages:
libglade2-dev

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 168/520
 make menuconfig

make menuconfig
▶ Useful when no graphics are available.
Pretty convenient too!
▶ Same interface found in other tools:
BusyBox, Buildroot...
▶ Required Debian packages:
libncurses-dev

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 169/520
 make nconfig

make nconfig
▶ A newer, similar text interface
▶ More user friendly (for example, easier
to access help information).
▶ However, lacking the shortcuts that
menuconfig offers in search results.
Therefore, much less convenient than
menuconfig.
▶ Required Debian packages:
libncurses-dev

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 170/520
 make oldconfig

make oldconfig
▶ Needed very often!
▶ Useful to upgrade a .config file from an earlier kernel release
▶ Asks for values for new parameters.
▶ ... unlike make menuconfig and make xconfig which silently set default values
for new parameters.
If you edit a .config file by hand, it’s useful to run make oldconfig afterwards, to set
values to new parameters that could have appeared because of dependency changes.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 171/520
 Undoing configuration changes

A frequent problem:
▶ After changing several kernel configuration settings, your kernel no longer works.
▶ If you don’t remember all the changes you made, you can get back to your
previous configuration:
$ cp .config.old .config
▶ All the configuration interfaces of the kernel (xconfig, menuconfig,
oldconfig...) keep this .config.old backup copy.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 172/520
 Linux kernel introduction

Compiling and installing the kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 173/520
 Choose a compiler

The compiler invoked by the kernel Makefile is $(CROSS_COMPILE)gcc

▶ When compiling natively
▶ Leave CROSS_COMPILE undefined and the kernel will be natively compiled for the host
architecture using gcc.
▶ When using a cross-compiler
▶ To make the difference with a native compiler, cross-compiler executables are
prefixed by the name of the target system, architecture and sometimes library.
Examples:
mips-linux-gcc: the prefix is mips-linux-
arm-linux-gnueabi-gcc: the prefix is arm-linux-gnueabi-
▶ So, you can specify your cross-compiler as follows:
export CROSS_COMPILE=arm-linux-gnueabi-
CROSS_COMPILE is actually the prefix of the cross compiling tools (gcc, as, ld,
objcopy, strip...).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 174/520
 Specifying ARCH and CROSS_COMPILE

There are actually two ways of defining ARCH and CROSS_COMPILE:

▶ Pass ARCH and CROSS_COMPILE on the make command line:
make ARCH=arm CROSS_COMPILE=arm-linux- ...
Drawback: it is easy to forget to pass these variables when you run any make
command, causing your build and configuration to be screwed up.
▶ Define ARCH and CROSS_COMPILE as environment variables:
export ARCH=arm
export CROSS_COMPILE=arm-linux-
Drawback: it only works inside the current shell or terminal. You could put these
settings in a file that you source every time you start working on the project. If
you only work on a single architecture with always the same toolchain, you could
even put these settings in your ~/.bashrc file to make them permanent and
visible from any terminal.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 175/520
 Kernel compilation

▶ make
▶ In the main kernel source directory!
▶ Remember to run multiple jobs in parallel if you have multiple CPU cores. Example:
make -j 8
▶ No need to run as root!
▶ Generates
▶ vmlinux, the raw uncompressed kernel image, in the ELF format, useful for
debugging purposes, but cannot be booted
▶ arch/<arch>/boot/*Image, the final, usually compressed, kernel image that can be
booted
▶ bzImage for x86, zImage for ARM, vmlinux.bin.gz for ARC, etc.
▶ arch/<arch>/boot/dts/*.dtb, compiled Device Tree files (on some architectures)
▶ All kernel modules, spread over the kernel source tree, as .ko (Kernel Object) files.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 176/520
 Kernel installation: native case

▶ make install
▶ Does the installation for the host system by default, so needs to be run as root.
▶ Installs
▶ /boot/vmlinuz-<version>
Compressed kernel image. Same as the one in arch/<arch>/boot
▶ /boot/System.map-<version>
Stores kernel symbol addresses for debugging purposes (obsolete: such information is
usually stored in the kernel itself)
▶ /boot/config-<version>
Kernel configuration for this version
▶ In GNU/Linux distributions, typically re-runs the bootloader configuration utility
to make the new kernel available at the next boot.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 177/520
 Kernel installation: embedded case

▶ make install is rarely used in embedded development, as the kernel image is a

single file, easy to handle.
▶ Another reason is that there is no standard way to deploy and use the kernel
image.
▶ Therefore making the kernel image available to the target is usually manual or
done through scripts in build systems.
▶ It is however possible to customize the make install behavior in
arch/<arch>/boot/install.sh

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 178/520
 Module installation: native case

▶ make modules_install
▶ Does the installation for the host system by default, so needs to be run as root
▶ Installs all modules in /lib/modules/<version>/
▶ kernel/
Module .ko (Kernel Object) files, in the same directory structure as in the sources.
▶ modules.alias, modules.aliases.bin
Aliases for module loading utilities. Used to find drivers for devices. Example line:
alias usb:v066Bp20F9d*dc*dsc*dp*ic*isc*ip*in* asix
▶ modules.dep, modules.dep.bin
Module dependencies
▶ modules.symbols, modules.symbols.bin
Tells which module a given symbol belongs to.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 179/520
 Module installation: embedded case

▶ In embedded development, you can’t directly use make modules_install as it

would install target modules in /lib/modules on the host!
▶ The INSTALL_MOD_PATH variable is needed to generate the module related files
and install the modules in the target root filesystem instead of your host root
filesystem:
make INSTALL_MOD_PATH=<dir>/ modules_install

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 180/520
 Kernel cleanup targets

▶ Clean-up generated files (to force re-compilation):

make clean
▶ Remove all generated files. Needed when switching from one
architecture to another. Caution: it also removes your .config
file!
make mrproper
▶ Also remove editor backup and patch reject files (mainly to
generate patches):
make distclean
▶ If you are in a git tree, remove all files not tracked (and ignored)
by git:
git clean -fdx

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 181/520
 Kernel building overview

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 182/520
 Linux kernel introduction

Booting the kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 183/520
 Device Tree (DT)

▶ Many embedded architectures have a lot of non-discoverable hardware.

▶ Depending on the architecture, such hardware is either described in BIOS ACPI
tables (x86), using C code directly within the kernel, or using a special hardware
description language in a Device Tree.
▶ The DT was created for PowerPC, and later was adopted by other architectures
(ARM, ARC...). Now Linux has DT support in most architectures, at least for
specific systems (for example for the OLPC on x86).
▶ A Device Tree Source, written by kernel developers, is compiled into a binary
Device Tree Blob, and needs to be passed to the kernel at boot time.
▶ There is one different Device Tree for each board/platform supported by the kernel,
available in arch/arm/boot/dts/<board>.dtb.
▶ The bootloader must load both the kernel image and the Device Tree Blob in
memory before starting the kernel.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 184/520
 Customize your board device tree!

Often needed for embedded board users:

▶ To describe external devices attached to
non-discoverable busses (such as I2C) and configure
them.
▶ To configure pin muxing: choosing what SoC
signals are made available on the board external
connectors.
▶ To configure some system parameters: flash
partitions, kernel command line (other ways exist)
▶ Useful reference: Device Tree for Dummies,
Thomas Petazzoni (Apr. 2014):
https://j.mp/1jQU6NR

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 185/520
 Booting with U-Boot

▶ Recent versions of U-Boot can boot the zImage binary.

▶ Older versions require a special kernel image format: uImage
▶ uImage is generated from zImage using the mkimage tool. It is done automatically
by the kernel make uImage target.
▶ On some ARM platforms, make uImage requires passing a LOADADDR environment
variable, which indicates at which physical memory address the kernel will be
executed.
▶ In addition to the kernel image, U-Boot can also pass a Device Tree Blob to the
kernel.
▶ The typical boot process is therefore:
1. Load zImage or uImage at address X in memory
2. Load <board>.dtb at address Y in memory
3. Start the kernel with bootz X - Y (zImage case), or bootm X - Y (uImage case)
The - in the middle indicates no initramfs

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 186/520
 Kernel command line

▶ In addition to the compile time configuration, the kernel behavior can be adjusted
with no recompilation using the kernel command line
▶ The kernel command line is a string that defines various arguments to the kernel
▶ It is very important for system configuration
▶ root= for the root filesystem (covered later)
▶ console= for the destination of kernel messages
▶ Many more exist. The most important ones are documented in
admin-guide/kernel-parameters in kernel documentation.
▶ This kernel command line can be, in order of priority (highest to lowest):
▶ Passed by the bootloader. In U-Boot, the contents of the bootargs environment
variable is automatically passed to the kernel.
▶ Specified in the Device Tree (for architectures which use it)
▶ Built into the kernel, using the CONFIG_CMDLINE option.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 187/520
 Practical lab - Kernel cross-compiling

▶ Set up the cross-compiling environment

▶ Configure and cross-compile the kernel for an
arm platform
▶ On this platform, interact with the bootloader
and boot your kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 188/520
 Linux kernel introduction

Using kernel modules

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 189/520
 Advantages of modules

▶ Modules make it easy to develop drivers without rebooting: load, test, unload,
rebuild, load...
▶ Useful to keep the kernel image size to the minimum (essential in GNU/Linux
distributions for PCs).
▶ Also useful to reduce boot time: you don’t spend time initializing devices and
kernel features that you only need later.
▶ Caution: once loaded, have full control and privileges in the system. No particular
protection. That’s why only the root user can load and unload modules.
▶ To increase security, possibility to allow only signed modules, or to disable module
support entirely.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 190/520
 Module dependencies

▶ Some kernel modules can depend on other modules, which need to be loaded first.
▶ Example: the ubifs module depends on the ubi and mtd modules.
▶ Dependencies are described both in
/lib/modules/<kernel-version>/modules.dep and in
/lib/modules/<kernel-version>/modules.dep.bin (binary hashed format)
These files are generated when you run make modules_install.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 191/520
 Kernel log

When a new module is loaded, related information is available in the kernel log.
▶ The kernel keeps its messages in a circular buffer (so that it doesn’t consume
more memory with many messages)
▶ Kernel log messages are available through the dmesg command (diagnostic
message)
▶ Kernel log messages are also displayed in the system console (console messages
can be filtered by level using the loglevel kernel command line parameter, or
completely disabled with the quiet parameter). Example:
console=ttyS0 root=/dev/mmcblk0p2 loglevel=5
▶ Note that you can write to the kernel log from user space too:
echo "<n>Debug info" > /dev/kmsg

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 192/520
 Module utilities (1)

<module_name>: name of the module file without the trailing .ko

▶ modinfo <module_name> (for modules in /lib/modules)
modinfo <module_path>.ko
Gets information about a module without loading it: parameters, license,
description and dependencies.
▶ sudo insmod <module_path>.ko
Tries to load the given module. The full path to the module object file must be
given.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 193/520
 Understanding module loading issues

▶ When loading a module fails, insmod often doesn’t give you enough details!
▶ Details are often available in the kernel log.
▶ Example:
$ sudo insmod ./intr_monitor.ko
insmod: error inserting './intr_monitor.ko': -1 Device or resource busy
$ dmesg
[17549774.552000] Failed to register handler for irq channel 2

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 194/520
 Module utilities (2)

▶ sudo modprobe <module_name>

Most common usage of modprobe: tries to load all the modules the given module
depends on, and then this module. Lots of other options are available. modprobe
automatically looks in /lib/modules/<version>/ for the object file
corresponding to the given module name.
▶ lsmod
Displays the list of loaded modules
Compare its output with the contents of /proc/modules!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 195/520
 Module utilities (3)

▶ sudo rmmod <module_name>

Tries to remove the given module.
Will only be allowed if the module is no longer in use (for example, no more
processes opening a device file)
▶ sudo modprobe -r <module_name>
Tries to remove the given module and all dependent modules (which are no longer
needed after removing the module)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 196/520
 Passing parameters to modules

▶ Find available parameters:

modinfo usb-storage
▶ Through insmod:
sudo insmod ./usb-storage.ko delay_use=0
▶ Through modprobe:
Set parameters in /etc/modprobe.conf or in any file in /etc/modprobe.d/:
options usb-storage delay_use=0
▶ Through the kernel command line, when the driver is built statically into the
kernel:
usb-storage.delay_use=0
▶ usb-storage is the driver name
▶ delay_use is the driver parameter name. It specifies a delay before accessing a USB
storage device (useful for rotating devices).
▶ 0 is the driver parameter value

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 197/520
 Check module parameter values

How to find/edit the current values for the parameters of a loaded module?
▶ Check /sys/module/<name>/parameters.
▶ There is one file per parameter, containing the parameter value.
▶ Also possible to change parameter values if these files have write permissions
(depends on the module code).
▶ Example:
echo 0 > /sys/module/usb_storage/parameters/delay_use

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 198/520
 Useful reading

Linux Kernel in a Nutshell, Dec. 2006

▶ By Greg Kroah-Hartman, O’Reilly
http://www.kroah.com/lkn/
▶ A good reference book and guide on configuring,
compiling and managing the Linux kernel sources.
▶ Freely available on-line!
Great companion to the printed book for easy electronic
searches!
Available as single PDF file on
https://bootlin.com/community/kernel/lkn/
▶ Getting old but still containing useful content.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 199/520
 Linux Root Filesystem

Linux Root Filesystem

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 200/520
 Linux Root Filesystem

Principle and solutions

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 201/520
 Filesystems

▶ Filesystems are used to organize data in directories and files on storage devices or
on the network. The directories and files are organized as a hierarchy
▶ In UNIX systems, applications and users see a single global hierarchy of files and
directories, which can be composed of several filesystems.
▶ Filesystems are mounted in a specific location in this hierarchy of directories
▶ When a filesystem is mounted in a directory (called mount point), the contents of
this directory reflects the contents of the storage device
▶ When the filesystem is unmounted, the mount point is empty again.
▶ This allows applications to access files and directories easily, regardless of their
exact storage location

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 202/520
 Filesystems (2)

▶ Create a mount point, which is just a directory

$ sudo mkdir /mnt/usbkey
▶ It is empty
$ ls /mnt/usbkey
$
▶ Mount a storage device in this mount point
$ sudo mount -t vfat /dev/sda1 /mnt/usbkey
$
▶ You can access the contents of the USB key
$ ls /mnt/usbkey
docs prog.c picture.png movie.avi
$

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 203/520
 mount / umount

▶ mount allows to mount filesystems

▶ mount -t type device mountpoint
▶ type is the type of filesystem (optional for non-virtual filesystems)
▶ device is the storage device, or network location to mount
▶ mountpoint is the directory where files of the storage device or network location will
be accessible
▶ mount with no arguments shows the currently mounted filesystems
▶ umount allows to unmount filesystems
▶ This is needed before rebooting, or before unplugging a USB key, because the Linux
kernel caches writes in memory to increase performance. umount makes sure that
these writes are committed to the storage.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 204/520
 Root filesystem

▶ A particular filesystem is mounted at the root of the hierarchy, identified by /

▶ This filesystem is called the root filesystem
▶ As mount and umount are programs, they are files inside a filesystem.
▶ They are not accessible before mounting at least one filesystem.
▶ As the root filesystem is the first mounted filesystem, it cannot be mounted with
the normal mount command
▶ It is mounted directly by the kernel, according to the root= kernel option
▶ When no root filesystem is available, the kernel panics:
Please append a correct "root=" boot option
Kernel panic - not syncing: VFS: Unable to mount root fs on unknown block(0,0)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 205/520
 Location of the root filesystem

▶ It can be mounted from different locations

▶ From the partition of a hard disk
▶ From the partition of a USB key
▶ From the partition of an SD card
▶ From the partition of a NAND flash chip or similar type of storage device
▶ From the network, using the NFS protocol
▶ From memory, using a pre-loaded filesystem (by the bootloader)
▶ etc.
▶ It is up to the system designer to choose the configuration for the system, and
configure the kernel behavior with root=

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 206/520
 Mounting rootfs from storage devices

▶ Partitions of a hard disk or USB key

▶ root=/dev/sdXY, where X is a letter indicating the device, and Y a number
indicating the partition
▶ /dev/sdb2 is the second partition of the second disk drive (either USB key or ATA
hard drive)
▶ Partitions of an SD card
▶ root=/dev/mmcblkXpY, where X is a number indicating the device and Y a number
indicating the partition
▶ /dev/mmcblk0p2 is the second partition of the first device
▶ Partitions of flash storage
▶ root=/dev/mtdblockX, where X is the partition number
▶ /dev/mtdblock3 is the fourth partition of a NAND flash chip (if only one NAND
flash chip is present)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 207/520
 Mounting rootfs over the network (1)
Once networking works, your root filesystem could be a directory on your GNU/Linux
development host, exported by NFS (Network File System). This is very convenient for
system development:
▶ Makes it very easy to update files on the root filesystem, without rebooting.
Much faster than through the serial port.
▶ Can have a big root filesystem even if you don’t have support for internal or
external storage yet.
▶ The root filesystem can be huge. You can even build native compiler tools and
build all the tools you need on the target itself (better to cross-compile though).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 208/520
 Mounting rootfs over the network (2)

On the development workstation side, a NFS server is needed

▶ Install an NFS server (example: Debian, Ubuntu)
sudo apt install nfs-kernel-server
▶ Add the exported directory to your /etc/exports file:
/home/tux/rootfs 192.168.1.111(rw,no_root_squash,no_subtree_check)
▶ 192.168.1.111 is the client IP address
▶ rw,no_root_squash,no_subtree_check are the NFS server options for this
directory export.
▶ Start or restart your NFS server (example: Debian, Ubuntu)
sudo /etc/init.d/nfs-kernel-server restart

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 209/520
 Mounting rootfs over the network (3)

▶ On the target system

▶ The kernel must be compiled with
▶ CONFIG_NFS_FS=y (NFS client support)
▶ CONFIG_IP_PNP=y (configure IP at boot time)
▶ CONFIG_ROOT_NFS=y (support for NFS as rootfs)
▶ The kernel must be booted with the following parameters:
▶ root=/dev/nfs (we want rootfs over NFS)
▶ ip=192.168.1.111 (target IP address)
▶ nfsroot=192.168.1.110:/home/tux/rootfs/ (NFS server details)
▶ You may need to add ”,nfsvers=3,tcp” to the nfsroot setting, as an NFS version
2 client and UDP may be rejected by the NFS server in recent GNU/Linux
distributions.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 210/520
 Mounting rootfs over the network (4)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 211/520
 rootfs in memory: initramfs (1)

▶ It is also possible to have the root filesystem integrated into the kernel image
▶ It is therefore loaded into memory together with the kernel
▶ This mechanism is called initramfs
▶ It integrates a compressed archive of the filesystem into the kernel image
▶ Variant: the compressed archive can also be loaded separately by the bootloader.
▶ It is useful for two cases
▶ Fast booting of very small root filesystems. As the filesystem is completely loaded at
boot time, application startup is very fast.
▶ As an intermediate step before switching to a real root filesystem, located on devices
for which drivers not part of the kernel image are needed (storage drivers, filesystem
drivers, network drivers). This is always used on the kernel of desktop/server
distributions to keep the kernel image size reasonable.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 212/520
 rootfs in memory: initramfs (2)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 213/520
 rootfs in memory: initramfs (3)

▶ The contents of an initramfs are defined at the kernel configuration level, with the
CONFIG_INITRAMFS_SOURCE option
▶ Can be the path to a directory containing the root filesystem contents
▶ Can be the path to a cpio archive
▶ Can be a text file describing the contents of the initramfs
(see documentation for details)
▶ The kernel build process will automatically take the contents of the
CONFIG_INITRAMFS_SOURCE option and integrate the root filesystem into the
kernel image
▶ Details (in kernel sources):
Documentation/filesystems/ramfs-rootfs-initramfs.txt
Documentation/early-userspace/README

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 214/520
 Linux Root Filesystem

Contents

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 215/520
 Root filesystem organization

▶ The organization of a Linux root filesystem in terms of directories is well-defined

by the Filesystem Hierarchy Standard
▶ https://wiki.linuxfoundation.org/lsb/fhs
▶ Most Linux systems conform to this specification
▶ Applications expect this organization
▶ It makes it easier for developers and users as the filesystem organization is similar in
all systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 216/520
 Important directories (1)

/bin Basic programs

/boot Kernel image (only when the kernel is loaded from a filesystem, not
common on non-x86 architectures)
/dev Device files (covered later)
/etc System-wide configuration
/home Directory for the users home directories
/lib Basic libraries
/media Mount points for removable media
/mnt Mount points for static media
/proc Mount point for the proc virtual filesystem

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 217/520
 Important directories (2)

/root Home directory of the root user

/sbin Basic system programs
/sys Mount point of the sysfs virtual filesystem
/tmp Temporary files
/usr /usr/bin Non-basic programs
/usr/lib Non-basic libraries
/usr/sbin Non-basic system programs
/var Variable data files. This includes spool directories and files,
administrative and logging data, and transient and temporary files

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 218/520
 Separation of programs and libraries

▶ Basic programs are installed in /bin and /sbin and basic libraries in /lib
▶ All other programs are installed in /usr/bin and /usr/sbin and all other libraries
in /usr/lib
▶ In the past, on UNIX systems, /usr was very often mounted over the network,
through NFS
▶ In order to allow the system to boot when the network was down, some binaries
and libraries are stored in /bin, /sbin and /lib
▶ /bin and /sbin contain programs like ls, ifconfig, cp, bash, etc.
▶ /lib contains the C library and sometimes a few other basic libraries
▶ All other programs and libraries are in /usr

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 219/520
 Linux Root Filesystem

Device Files

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 220/520
 Devices

▶ One of the kernel important roles is to allow applications to access hardware

devices
▶ In the Linux kernel, most devices are presented to user space applications through
two different abstractions
▶ Character device
▶ Block device
▶ Internally, the kernel identifies each device by a triplet of information
▶ Type (character or block)
▶ Major (typically the category of device)
▶ Minor (typically the identifier of the device)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 221/520
 Types of devices

▶ Block devices
▶ A device composed of fixed-sized blocks, that can be read and written to store data
▶ Used for hard disks, USB keys, SD cards, etc.
▶ Character devices
▶ Originally, an infinite stream of bytes, with no beginning, no end, no size. The pure
example: a serial port.
▶ Used for serial ports, terminals, but also sound cards, video acquisition devices,
frame buffers
▶ Most of the devices that are not block devices are represented as character devices
by the Linux kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 222/520
 Devices: everything is a file

▶ A very important UNIX design decision was to represent most system objects as
files
▶ It allows applications to manipulate all system objects with the normal file API
(open, read, write, close, etc.)
▶ So, devices had to be represented as files to the applications
▶ This is done through a special artifact called a device file
▶ It is a special type of file, that associates a file name visible to user space
applications to the triplet (type, major, minor) that the kernel understands
▶ All device files are by convention stored in the /dev directory

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 223/520
 Device files examples

Example of device files in a Linux system

$ ls -l /dev/ttyS0 /dev/tty1 /dev/sda1 /dev/sda2 /dev/zero

brw-rw---- 1 root disk 8, 1 2011-05-27 08:56 /dev/sda1
brw-rw---- 1 root disk 8, 2 2011-05-27 08:56 /dev/sda2
crw------- 1 root root 4, 1 2011-05-27 08:57 /dev/tty1
crw-rw---- 1 root dialout 4, 64 2011-05-27 08:56 /dev/ttyS0
crw-rw-rw- 1 root root 1, 5 2011-05-27 08:56 /dev/zero

Example C code that uses the usual file API to write data to a serial port
int fd;
fd = open("/dev/ttyS0", O_RDWR);
write(fd, "Hello", 5);
close(fd);

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 224/520
 Creating device files

▶ Before Linux 2.6.32, on basic Linux systems, the device files had to be created
manually using the mknod command
▶ mknod /dev/<device> [c|b] major minor
▶ Needed root privileges
▶ Coherency between device files and devices handled by the kernel was left to the
system developer
▶ The devtmpfs virtual filesystem can be mounted on /dev and contains all the
devices known to the kernel. The CONFIG_DEVTMPFS_MOUNT kernel configuration
option makes the kernel mount it automatically at boot time, except when
booting on an initramfs.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 225/520
 Linux Root Filesystem

Pseudo Filesystems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 226/520
 proc virtual filesystem

▶ The proc virtual filesystem exists since the beginning of Linux

▶ It allows
▶ The kernel to expose statistics about running processes in the system
▶ The user to adjust at runtime various system parameters about process
management, memory management, etc.
▶ The proc filesystem is used by many standard user space applications, and they
expect it to be mounted in /proc
▶ Applications such as ps or top would not work without the proc filesystem
▶ Command to mount proc:
mount -t proc nodev /proc
▶ Documentation/filesystems/proc.txt in the kernel sources
▶ man proc

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 227/520
 proc contents

▶ One directory for each running process in the system

▶ /proc/<pid>
▶ cat /proc/3840/cmdline
▶ It contains details about the files opened by the process, the CPU and memory
usage, etc.
▶ /proc/interrupts, /proc/devices, /proc/iomem, /proc/ioports contain
general device-related information
▶ /proc/cmdline contains the kernel command line
▶ /proc/sys contains many files that can be written to adjust kernel parameters
▶ They are called sysctl. See Documentation/sysctl/ in kernel sources.
▶ Example
echo 3 > /proc/sys/vm/drop_caches

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 228/520
 sysfs filesystem

▶ It allows to represent in user space the vision that the kernel has of the buses,
devices and drivers in the system
▶ It is useful for various user space applications that need to list and query the
available hardware, for example udev or mdev.
▶ All applications using sysfs expect it to be mounted in the /sys directory
▶ Command to mount /sys:
mount -t sysfs nodev /sys
▶ $ ls /sys/
block bus class dev devices firmware
fs kernel module power

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 229/520
 Linux Root Filesystem

Minimal filesystem

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 230/520
 Basic applications

▶ In order to work, a Linux system needs at least a few applications

▶ An init application, which is the first user space application started by the kernel after
mounting the root filesystem (see https://en.wikipedia.org/wiki/Init):
▶ The kernel tries to run the command specified by the init= command line
parameter if available.
▶ Otherwise, it tries to run /sbin/init, /bin/init, /etc/init and /bin/sh.
▶ In the case of an initramfs, it will only look for /init. Another path can be supplied
by the rdinit kernel argument.
▶ If none of this works, the kernel panics and the boot process is stopped.
▶ The init application is responsible for starting all other user space applications and
services:
▶ A shell, to implement scripts, automate tasks, and allow a user to interact with the system
▶ Basic UNIX executables, for use in system scripts or in interactive shells: mv, cp, mkdir,
cat, modprobe, mount, ifconfig, etc.
▶ These basic components have to be integrated into the root filesystem to make it usable

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 231/520
 Overall booting process

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 232/520
 Overall booting process with initramfs

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 233/520
 Busybox

Busybox

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 234/520
 Why Busybox?

▶ A Linux system needs a basic set of programs to work

▶ An init program
▶ A shell
▶ Various basic utilities for file manipulation and system configuration
▶ In normal Linux systems, these programs are provided by different projects
▶ coreutils, bash, grep, sed, tar, wget, modutils, etc. are all different projects
▶ A lot of different components to integrate
▶ Components not designed with embedded systems constraints in mind: they are not
very configurable and have a wide range of features
▶ Busybox is an alternative solution, extremely common on embedded systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 235/520
 General purpose toolbox: BusyBox

▶ Rewrite of many useful UNIX command line utilities

▶ Created in 1995 to implement a rescue and installer system for Debian, fitting in a
single floppy disk.
▶ Integrated into a single project, which makes it easy to work with
▶ Designed with embedded systems in mind: highly configurable, no unnecessary
features
▶ All the utilities are compiled into a single executable, /bin/busybox
▶ Symbolic links to /bin/busybox are created for each application integrated into
Busybox
▶ For a fairly featureful configuration, less than 500 KB (statically compiled with
uClibc) or less than 1 MB (statically compiled with glibc).
▶ https://www.busybox.net/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 236/520
 BusyBox commands!

[, [[, acpid, add-shell, addgroup, adduser, adjtimex, ar, arch, arp, arping, awk, base64, basename, bbconfig, bc, beep,
blkdiscard, blkid, blockdev, bootchartd, brctl, bunzip2, busybox, bzcat, bzip2, cal, cat, chat, chattr, chcon, chgrp, chmod,
chown, chpasswd, chpst, chroot, chrt, chvt, cksum, clear, cmp, comm, conspy, cp, cpio, crond, crontab, cryptpw, cttyhack, cut,
date, dc, dd, deallocvt, delgroup, deluser, depmod, devmem, df, diff, dirname, dmesg, dnsd, dnsdomainname, dos2unix, dpkg, dpkg-
deb, du, dumpkmap, dumpleases, echo, ed, egrep, eject, env, envdir, envuidgid, ether-wake, expand, expr, factor, fakeidentd,
fallocate, false, fatattr, fbset, fbsplash, fdflush, fdformat, fdisk, fgconsole, fgrep, find, findfs, flash_eraseall, flash_lock,
flash_unlock, flashcp, flock, fold, free, freeramdisk, fsck, fsck.minix, fsfreeze, fstrim, fsync, ftpd, ftpget, ftpput, fuser,
getenforce, getopt, getsebool, getty, grep, groups, gunzip, gzip, halt, hd, hdparm, head, hexdump, hexedit, hostid, hostname,
httpd, hush, hwclock, id, ifconfig, ifdown, ifenslave, ifplugd, ifup, inetd, init, insmod, install, ionice, iostat, ip, ipaddr,
ipcalc, ipcrm, ipcs, iplink, ipneigh, iproute, iprule, iptunnel, kbd_mode, kill, killall, killall5, klogd, last, less, link,
linux32, linux64, linuxrc, ln, load_policy, loadfont, loadkmap, logger, login, logname, logread, losetup, lpd, lpq, lpr, ls,
lsattr, lsmod, lsof, lspci, lsscsi, lsusb, lzcat, lzma, lzop, lzopcat, makedevs, makemime, man, matchpathcon, md5sum, mdev, mesg,
microcom, minips, mkdir, mkdosfs, mke2fs, mkfifo, mkfs.ext2, mkfs.minix, mkfs.reiser, mkfs.vfat, mknod, mkpasswd, mkswap, mktemp,
modinfo, modprobe, more, mount, mountpoint, mpstat, mt, mv, nameif, nanddump, nandwrite, nbd-
client, nc, netcat, netstat, nice, nl, nmeter, nohup, nologin, nproc, nsenter, nslookup, ntpd, nuke, od, openvt, partprobe,
passwd, paste, patch, pgrep, pidof, ping, ping6, pipe_progress, pivot_root, pkill, pmap, popmaildir, poweroff, printenv, printf,
ps, pscan, pstree, pwd, pwdx, raidautorun, rdate, rdev, readahead, readlink, readprofile, realpath, reboot, reformime, remove-
shell, renice, reset, resize, restorecon, resume, rev, rfkill, rm, rmdir, rmmod, route, rpm, rpm2cpio, rtcwake, run-init, run-
parts, runcon, runlevel, runsv, runsvdir, rx, script, scriptreplay, sed, selinuxenabled, sendmail, seq, sestatus, setarch,
setconsole, setenforce, setfattr, setfiles, setfont, setkeycodes, setlogcons, setpriv, setsebool, setserial, setsid, setuidgid,
sh, sha1sum, sha256sum, sha3sum, sha512sum, showkey, shred, shuf, slattach, sleep, smemcap, softlimit, sort, split, ssl_client,
start-stop-daemon, stat, strings, stty, su, sulogin, sum, sv, svc, svlogd, svok, swapoff, swapon, switch_root, sync, sysctl,
syslogd, tac, tail, tar, taskset, tc, tcpsvd, tee, telnet, telnetd, test, tftp, tftpd, time, timeout, top, touch, tr, traceroute,
traceroute6, true, truncate, ts, tty, ttysize, tunctl, tune2fs, ubiattach, ubidetach, ubimkvol, ubirename, ubirmvol, ubirsvol,
ubiupdatevol, udhcpc, udhcpd, udpsvd, uevent, umount, uname, uncompress, unexpand, uniq, unit, unix2dos, unlink, unlzma, unlzop,
unxz, unzip, uptime, users, usleep, uudecode, uuencode, vconfig, vi, vlock, volname, w, wall, watch, watchdog, wc, wget, which,
who, whoami, whois, xargs, xxd, xz, xzcat, yes, zcat, zcip

Source: run /bin/busybox

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 237/520
 Applet highlight: Busybox init

▶ Busybox provides an implementation of an init program

▶ Simpler than the init implementation found on desktop/server systems (SysV init
or systemd)
▶ A single configuration file: /etc/inittab
▶ Each line has the form <id>::<action>:<process>
▶ Allows to run services at startup, and to make sure that certain services are always
running on the system
▶ See examples/inittab in Busybox for details on the configuration

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 238/520
 Applet highlight - BusyBox vi

▶ If you are using BusyBox, adding vi support only adds 20K. (built with shared
libraries, using uClibc).
▶ You can select which exact features to compile in.
▶ Users hardly realize that they are using a lightweight vi version!
▶ Tip: you can learn vi on the desktop, by running the vimtutor command.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 239/520
 Configuring BusyBox

▶ Get the latest stable sources from https://busybox.net

▶ Configure BusyBox (creates a .config file):
▶ make defconfig
Good to begin with BusyBox.
Configures BusyBox with all options for regular users.
▶ make allnoconfig
Unselects all options. Good to configure only what you need.
▶ make xconfig (graphical, needs the libqt3-mt-dev package)
or make menuconfig (text)
Same configuration interfaces as the ones used by the Linux kernel (though older
versions are used).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 240/520
 BusyBox make xconfig

You can choose:

▶ the commands to
compile,
▶ and even the
command options
and features that you
need!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 241/520
 Compiling BusyBox

▶ Set the cross-compiler prefix in the configuration interface:

Settings -> Build Options -> Cross Compiler prefix
Example: arm-linux-
▶ Set the installation directory in the configuration interface:
Settings -> Installation Options -> BusyBox installation prefix
▶ Add the cross-compiler path to the PATH environment variable:
export PATH=$HOME/x-tools/arm-unknown-linux-uclibcgnueabi/bin:$PATH
▶ Compile BusyBox:
make
▶ Install it (this creates a UNIX directory structure symbolic links to the busybox
executable):
make install

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 242/520
 Practical lab - A tiny embedded system

▶ Make Linux boot on a directory on your

workstation, shared by NFS
▶ Create and configure a minimalistic Linux
embedded system
▶ Install and use BusyBox
▶ System startup with /sbin/init
▶ Set up a simple web interface
▶ Use shared libraries

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 243/520
 Block filesystems

Block filesystems

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 244/520
 Block filesystems

Block devices

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 245/520
 Block vs. flash

▶ Storage devices are classified in two main types: block devices and flash devices
▶ They are handled by different subsystems and different filesystems
▶ Block devices can be read and written to on a per-block basis, in random order,
without erasing.
▶ Hard disks, RAM disks
▶ USB keys, SSD, SD card, eMMC: these are based on flash storage, but have an
integrated controller that emulates a block device, managing the flash in a
transparent way.
▶ Raw flash devices are driven by a controller on the SoC. They can be read, but
writing requires prior erasing, and often occurs on a larger size than the “block”
size.
▶ NOR flash, NAND flash

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 246/520
 Block device list

▶ The list of all block devices available in the system can be found in
/proc/partitions
$ cat /proc/partitions
major minor #blocks name

179 0 3866624 mmcblk0

179 1 73712 mmcblk0p1
179 2 3792896 mmcblk0p2
8 0 976762584 sda
8 1 1060258 sda1
8 2 975699742 sda2
▶ /sys/block/ also stores information about each block device, for example
whether it is removable storage or not.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 247/520
 Partitioning

▶ Block devices can be partitioned to store different parts of a system

▶ The partition table is stored inside the device itself, and is read and analyzed
automatically by the Linux kernel
▶ mmcblk0 is the entire device
▶ mmcblk0p2 is the second partition of mmcblk0
▶ Two partition table formats:
▶ MBR, the legacy format
▶ GPT, the new format, not used everywhere yet, supporting disks bigger than 2 TB.
▶ Numerous tools to create and modify the partitions on a block device: fdisk,
cfdisk, sfdisk, parted, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 248/520
 Transfering data to a block device

▶ It is often necessary to transfer data to or from a block device in a raw way

▶ Especially to write a filesystem image to a block device
▶ This directly writes to the block device itself, bypassing any filesystem layer.
▶ The block devices in /dev/ allow such raw access
▶ dd (disk duplicate) is the tool of choice for such transfers:
▶ dd if=/dev/mmcblk0p1 of=testfile bs=1M count=16
Transfers 16 blocks of 1 MB from /dev/mmcblk0p1 to testfile
▶ dd if=testfile of=/dev/sda2 bs=1M seek=4
Transfers the complete contents of testfile to /dev/sda2, by blocks of 1 MB, but
starting at offset 4 MB in /dev/sda2
▶ Typical mistake: copying a file to a filesystem without mounting it first:
dd if=zImage of=/dev/sda1
Instead, you should use:
sudo mount /dev/sda1 /boot
cp zImage /boot/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 249/520
 Block filesystems

Available filesystems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 250/520
 Standard Linux filesystem format: ext2, ext3, ext4

▶ The standard filesystem used on Linux systems is the series of ext{2,3,4}

filesystems
▶ ext2
▶ ext3, brought journaling compared to ext2
▶ ext4, mainly brought performance improvements and support for even larger
filesystems
▶ It supports all features Linux needs in a root filesystem: permissions, ownership,
device files, symbolic links, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 251/520
 Journaled filesystems

▶ Unlike simpler filesystems (ext2, vfat...),

designed to stay in a coherent state even after
system crashes or a sudden poweroff.
▶ Writes are first described in the journal before
being committed to files (can be all writes, or
only metadata writes depending on the
configuration)
▶ Allows to skip a full disk check at boot time
after an unclean shutdown.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 252/520
 Filesystem recovery after crashes

▶ Thanks to the journal, the recovery at boot

time is quick, since the operations in progress
at the moment of the unclean shutdown are
clearly identified
▶ Does not mean that the latest writes made it
to the storage: this depends on syncing the
changes to the filesystem.
See https://en.wikipedia.org/wiki/
Journaling_file_system for further details.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 253/520
 Other journaled Linux/UNIX filesystems

▶ btrfs, intended to become the next standard filesystem for Linux. Integrates
numerous features: data checksuming, integrated volume management,
snapshots, etc.
▶ XFS, high-performance filesystem inherited from SGI IRIX, still actively developed.
▶ JFS, inherited from IBM AIX. No longer actively developed, provided mainly for
compatibility.
▶ reiserFS, used to be a popular filesystem, but its latest version Reiser4 was
never merged upstream.
All those filesystems provide the necessary functionalities for Linux systems: symbolic
links, permissions, ownership, device files, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 254/520
 F2FS: filesystem for flash-based storage

https://en.wikipedia.org/wiki/F2FS
▶ Filesystem that takes into account the characteristics of flash-based storage:
eMMC, SD cards, SSD, etc.
▶ Developed and contributed by Samsung
▶ Available in the mainline Linux kernel
▶ For optimal results, need a number of details about the storage internal behavior
which may not easy to get
▶ Benchmarks: best performer on flash devices most of the time:
See https://lwn.net/Articles/520003/
▶ Technical details: https://lwn.net/Articles/518988/
▶ Not as widely used as ext3,4, even on flash-based storage.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 255/520
 Squashfs: read-only filesystem

▶ Read-only, compressed filesystem for block devices. Fine for parts of a filesystem
which can be read-only (kernel, binaries...)
▶ Great compression rate, which generally brings improved read performance
▶ Used in most live CDs and live USB distributions
▶ Supports several compression algorithms (LZO, XZ, etc.)
▶ Benchmarks: roughly 3 times smaller than ext3, and 2-4 times faster
(https://elinux.org/Squash_Fs_Comparisons)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 256/520
 Compatibility filesystems

Linux also supports several other filesystem formats, mainly to be interopable with
other operating systems:
▶ vfat for compatibility with the FAT filesystem used in the Windows world and on
numerous removable devices
▶ Also convenient to store bootloader binaries (FAT easy to understand for ROM code)
▶ This filesystem does not support features like permissions, ownership, symbolic links,
etc. Cannot be used for a Linux root filesystem.
▶ Linux now supports the exFAT filesystem too (exfat).
▶ ntfs for compatibility with the NTFS filesystem used on Windows
▶ hfs for compatibility with the HFS filesystem used on Mac OS
▶ iso9660, the filesystem format used on CD-ROMs, obviously a read-only
filesystem

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 257/520
 tmpfs: filesystem in RAM

▶ Not a block filesystem of course!

▶ Perfect to store temporary data in RAM: system log files, connection data,
temporary files...
▶ More space-efficient than ramdisks: files are directly in the file cache, grows and
shrinks to accommodate stored files
▶ How to use: choose a name to distinguish the various tmpfs instances you could
have. Examples:
mount -t tmpfs run /var/run
mount -t tmpfs shm /dev/shm
▶ See Documentation/filesystems/tmpfs.txt in kernel sources.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 258/520
 Block filesystems

Using block filesystems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 259/520
 Creating ext2/ext3/ext4 filesystems

▶ To create an empty ext2/ext3/ext4 filesystem on a block device or inside an

already-existing image file
▶ mkfs.ext2 /dev/hda3
▶ mkfs.ext3 /dev/sda2
▶ mkfs.ext4 /dev/sda3
▶ mkfs.ext2 disk.img
▶ To create a filesystem image from a directory containing all your files and
directories
▶ Use the genext2fs tool, from the package of the same name
▶ This tool only supports ext2. No equivalents exist for ext3 and ext4.
▶ genext2fs -d rootfs/ rootfs.img
▶ Your image is then ready to be transferred to your block device

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 260/520
 Mounting filesystem images

▶ Once a filesystem image has been created, one can access and modifies its
contents from the development workstation, using the loop mechanism
▶ Example:
genext2fs -d rootfs/ rootfs.img
mkdir /tmp/tst
mount -t ext2 -o loop rootfs.img /tmp/tst
▶ In the /tmp/tst directory, one can access and modify the contents of the
rootfs.img file.
▶ This is possible thanks to loop, which is a kernel driver that emulates a block
device with the contents of a file.
▶ Do not forget to run umount before using the filesystem image!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 261/520
 Creating squashfs filesystems

▶ Need to install the squashfs-tools package

▶ Can only create an image: creating an empty squashfs filesystem would be
useless, since it’s read-only.
▶ To create a squashfs image:
▶ mksquashfs rootfs/ rootfs.sqfs -noappend
▶ -noappend: re-create the image from scratch rather than appending to it
▶ Mounting a squashfs filesystem:
▶ mount -t squashfs /dev/<device> /mnt

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 262/520
 Mixing read-only and read-write filesystems

Good idea to split your block storage into:

▶ A compressed read-only partition (Squashfs)
Typically used for the root filesystem (binaries,
kernel...).
Compression saves space. Read-only access protects
your system from mistakes and data corruption.
▶ A read-write partition with a journaled filesystem (like
ext4)
Used to store user or configuration data.
Guarantees filesystem integrity after power off or
crashes.
▶ Ram storage for temporary files (tmpfs)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 263/520
 Issues with flash-based block storage

▶ Flash storage made available only through a block interface.

▶ Hence, no way to access a low level flash interface and use the Linux filesystems
doing wear leveling.
▶ No details about the layer (Flash Translation Layer) they use. Details are kept as
trade secrets, and may hide poor implementations.
▶ Not knowing about the wear leveling algorithm, it is highly recommended to limit
the number of writes to these devices.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 264/520
 Practical lab - Block filesystems

▶ Creating partitions on your block storage

▶ Booting your system with a mix of filesystems:
SquashFS for the root filesystem (including
applications), ext4 for configuration and user
data, and tmpfs for temporary system files.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 265/520
 Flash storage and filesystems

Flash storage and

filesystems

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 266/520
 Block devices vs flash devices: reminder

▶ Block devices:
▶ Allow for random data access using fixed size blocks
▶ Do not require special care when writing on the media
▶ Block size is relatively small (minimum 512 bytes, can be increased for performance
reasons)
▶ Considered as reliable (if the storage media is not, some hardware or software parts
are supposed to make it reliable)
▶ Flash devices:
▶ Allow for random data access too
▶ Require special care before writing on the media (erasing the region you are about to
write on)
▶ Erase, write and read operation might not use the same block size
▶ Reliability depends on the flash technology

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 267/520
 NAND flash chips: how they work?

▶ Encode bits with voltage levels

▶ Start with all bits set to 1
▶ Programming implies changing some bits from 1 to 0
▶ Restoring bits to 1 is done via the ERASE operation
▶ Programming and erasing is not done on a per bit or per byte basis
▶ Organization
▶ Page: minimum unit for PROGRAM operation
▶ Block: minimum unit for ERASE operation

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 268/520
 NAND flash storage: organization

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 269/520
 NAND flash storage: constraints
▶ Reliability
▶ Far less reliable than NOR flash
▶ Reliability depends on the NAND flash technology (SLC, MLC)
▶ Require additional mechanisms to recover from bit flips: ECC (Error Correcting
Code)
▶ ECC information stored in the OOB (Out-of-band area)
▶ Lifetime
▶ Short lifetime compared to other storage media
▶ Lifetime depends on the NAND flash technology (SLC, MLC): between 1000000 and
1000 erase cycles per block
▶ Wear leveling mechanisms are required
▶ Bad block detection/handling required too
▶ Despite the number of constraints brought by NAND they are widely used in
embedded systems for several reasons:
▶ Cheaper than other flash technologies
▶ Provide high capacity storage
▶ Provide good performance (both in read and write access)
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 270/520
 NAND flash: ECC

▶ ECC partly addresses the reliability problem on NAND flash

▶ Operates on blocks (usually 512 or 1024 bytes)
▶ ECC data are stored in the OOB area
▶ Three algorithms:
▶ Hamming: can fixup a single bit per block
▶ Reed-Solomon: can fixup several bits per block
▶ BCH: can fixup several bits per block
▶ BCH and Reed-Solomon strength depends on the size allocated for ECC data,
which in turn depends on the OOB size
▶ NAND manufacturers specify the required ECC strength in their datasheets:
ignoring these requirements might compromise data integrity

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 271/520
 The MTD subsystem (1)
▶ MTD stands for Memory Technology Devices
▶ Generic subsystem in Linux dealing with all types of storage media that are not
fitting in the block subsystem
▶ Supported media types: RAM, ROM, NOR flash, NAND flash, Dataflash
▶ Independent of the communication interface (drivers available for parallel, SPI,
direct memory mapping, ...)
▶ Abstract storage media characteristics and provide a simple API to access MTD
devices
▶ MTD device characteristics exposed to users:
▶ erasesize: minimum erase size unit
▶ writesize: minimum write size unit
▶ oobsize: extra size to store metadata or ECC data
▶ size: device size
▶ flags: information about device type and capabilities
▶ Various kinds of MTD ”users” in the kernel: file-systems, block device emulation
layers, user space interfaces...
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 272/520
 The MTD subsystem (2)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 273/520
 MTD partitioning

▶ MTD devices are usually partitioned

▶ It allows to use different areas of the flash for different purposes: read-only
filesystem, read-write filesystem, backup areas, bootloader area, kernel area, etc.
▶ Unlike block devices, which contains their own partition table, the partitioning of
MTD devices is described externally (don’t want to put it in a flash sector which
could become bad)
▶ Specified in the board Device Tree
▶ Hard-coded into the kernel code (if no Device Tree)
▶ Specified through the kernel command line
▶ Each partition becomes a separate MTD device
▶ Different from block device labeling (hda3, sda2)
▶ /dev/mtd1 is either the second partition of the first flash device, or the first partition
of the second flash device
▶ Note that the master MTD device (the device those partitions belongs to) is not
exposed in /dev

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 274/520
 Linux: definition of MTD partitions
The Device Tree is the standard place to define default MTD partitions for platforms with
Device Tree support.
Example from arch/arm/boot/dts/omap3-overo-base.dtsi:
nand@0,0 {
linux,mtd-name= "micron,mt29c4g96maz";
[...]
ti,nand-ecc-opt = "bch8"
[...]
partition@0 {
label = "SPL";
reg = <0 0x80000>; /* 512KiB */
};
partition@80000 {
label = "U-Boot";
reg = <0x80000 0x1C0000>; /* 1792KiB */
};
partition@1c0000 {
label = "Environment";
reg = <0x240000 0x40000>; /* 256KiB */
};
[...]
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 275/520
 U-Boot: defining MTD partitions (1)

▶ U-Boot allows to define MTD partitions on flash devices, using the same syntax
as Linux for declaring them.
▶ Sharing definitions allows to eliminate the risk of mismatches between Linux and
U-Boot.
▶ Named partitions are also easier to use, and much less error prone than using
offsets.
▶ Use flash specific commands (detailed soon), and pass partition names instead of
numerical offsets
▶ Example: nand erase.part <partname>

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 276/520
 U-Boot: defining MTD partitions (2)
▶ Example:
setenv mtdids nand0=omap2-nand.0
setenv mtdparts mtdparts=omap2-nand.0:512k(XLoader)ro,1536k(UBoot)ro,512k(Env),4m(Kernel),-(Root)

▶ This defines 5 partitions in the omap2-nand.0 device:

▶ 1st stage bootloader (512 KiB, read-only)
▶ U-Boot (1536 KiB, read-only)
▶ U-Boot environment (512 KiB)
▶ Kernel (4 MiB)
▶ Root filesystem (Remaining space)
▶ Partition sizes must be multiple of the erase block size. You can use sizes in
hexadecimal too. Remember the below sizes: 0x20000 = 128 KiB, 0x100000 = 1
MiB, 0x1000000 = 16 MiB
▶ ro lists the partition as read only
▶ - is used to use all the remaining space.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 277/520
 U-Boot: defining MTD partitions (3)

mtdids associates a U-Boot flash device name to a Linux flash device name:
setenv mtdids <devid>=<mtdid>[,<devid>=<mtdid>]
That’s required because the Linux name is used in partition definitions.
▶ devid: U-Boot device identifier (from nand info or flinfo)
▶ mtdid: Linux mtd identifier. Displayed when booting the Linux kernel:
NAND device: Manufacturer ID: 0x2c, Chip ID: 0xbc (Micron NAND 512MiB 1,8V 16-bit)
Creating 5 MTD partitions on "omap2-nand.0":
0x000000000000-0x000000080000 : "X-Loader"
0x000000080000-0x000000200000 : "U-Boot"
0x000000200000-0x000000280000 : "Environment"
0x000000280000-0x000000580000 : "Kernel"
0x000000580000-0x000020000000 : "File System"

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 278/520
 U-Boot: defining MTD partitions (4)

mtdparts defines partitions for the different devices

▶ setenv mtdparts mtdparts=<mtdid>:<partition>[,partition]
▶ partition format: <size>[@offset](<name>)[ro]
Use the mtdparts command to process the mtdids and mtdparts settings and activate
partitions in U-Boot.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 279/520
 U-Boot: sharing partition definitions with Linux

Here is a recommended way to pass partition definitions from U-Boot to Linux:

▶ Define a bootargs_base environment variable:
setenv bootargs_base console=ttyS0 root=....
▶ Define the final kernel command line (bootargs) through the bootcmd
environment variable: setenv bootcmd 'setenv bootargs ${bootargs_base}
${mtdparts}; <rest of bootcmd>'

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 280/520
 U-Boot: manipulating NAND devices
U-Boot provides a set of commands to manipulate NAND devices, grouped under the
nand command
▶ nand info
Show available NAND devices and characteristics
▶ nand device [dev]
Select or display the active NAND device
▶ nand read[.option] <addr> <offset|partname> <size>
Read data from NAND
▶ nand write[.option] <addr> <offset|partname> <size>
Write data on NAND
▶ Use nand write.trimffs to avoid writing empty pages (those filled with 0xff)
▶ nand erase <offset> <size>
Erase a NAND region
▶ nand erase.part <partname>
Erase a NAND partition
▶ More commands for debugging purposes
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 281/520
 Linux: MTD devices interface with user space

▶ MTD devices are visible in /proc/mtd

▶ User space only sees MTD partitions, not the flash device under those partitions
(unless the kernel is compiled with CONFIG_MTD_PARTITIONED_MASTER)
▶ The mtdchar driver creates a character device for each MTD device/partition of
the system
▶ Usually named /dev/mtdX or /dev/mtdXro
▶ Provide ioctl() to erase and manage the flash
▶ Used by the mtd-utils utilities

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 282/520
 Linux: user space flash management tools

▶ mtd-utils is a set of utilities to manipulate MTD devices

▶ mtdinfo to get detailed information about an MTD device
▶ flash_erase to partially or completely erase a given MTD device
▶ flashcp to write to NOR flash
▶ nandwrite to write to NAND flash
▶ Flash filesystem image creation tools: mkfs.jffs2, mkfs.ubifs, ubinize, etc.
▶ On your workstation: usually available as the mtd-utils package in your
distribution.
▶ On your embedded target: most commands now also available in BusyBox.
▶ See http://www.linux-mtd.infradead.org/.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 283/520
 Flash wear leveling (1)

▶ Wear leveling consists in distributing erases over the whole flash device to avoid
quickly reaching the maximum number of erase cycles on blocks that are written
really often
▶ Can be done in:
▶ the filesystem layer (JFFS2, YAFFS2, ...)
▶ an intermediate layer dedicated to wear leveling (UBI)
▶ The wear leveling implementation is what makes your flash lifetime good or not

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 284/520
 Flash wear leveling (2)

Flash users should also take the limited lifetime of flash devices into account by taking
additional precautions
▶ Do not use your flash storage as swap area (rare in embedded systems anyway)
▶ Mount your filesystems as read-only whenever possible.
▶ Keep volatile files in RAM (tmpfs)
▶ Don’t use the sync mount option (commits writes immediately). Use the fsync()
system call for per-file synchronization.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 285/520
 Flash file-systems

▶ ’Standard’ file systems are meant to work on block devices

▶ Specific file systems have been developed to deal flash constraints
▶ These file systems are relying on the MTD layer to access flash chips
▶ There are several legacy flash filesystems which might be useful for specific usage:
JFFS2, YAFFS2.
▶ Nowadays, UBI/UBIFS is the de facto standard for medium to large capacity
NANDs

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 286/520
 Legacy flash filesystems: JFFS2

Standard file
API
▶ Supports on the fly compression
▶ Wear leveling, power failure resistant
JFFS2
▶ Available in the official Linux kernel filesystem
▶ Boot time depends on the filesystem size: doesn’t scale
well for large partitions because needs to scan the whole
storage at boot time. Need to enable MTD
CONFIG_JFFS2_SUMMARY to address this issue. driver
▶ http://www.linux-
mtd.infradead.org/doc/jffs2.html

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 287/520
 Legacy flash filesystems: YAFFS2

Standard file
API

▶ Mainly supports NAND flash

▶ No compression YAFFS2
▶ Wear leveling, power failure resistant filesystem
▶ Fast boot time
▶ Not part of the official Linux kernel: code only available
MTD
separately (Dual GPL / Proprietary license for non
driver
Linux operating systems)
▶ https://yaffs.net/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 288/520
 UBI/UBIFS
Standard file
API

▶ Aimed at replacing JFFS2 by addressing its limitations UBIFS

▶ Design choices: filesystem
▶ Split the wear leveling and filesystem layers
▶ Add some flexibility UBI
▶ Focus on scalability, performance and reliability
▶ Drawback: introduces noticeable space overhead,
especially when used on small devices or partitions. MTD
driver
JFFS2 still makes sense on small MTD partitions.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 289/520
 UBI (1)

Unsorted Block Images

▶ http://www.linux-mtd.infradead.org/doc/ubi.html
▶ Volume management system on top of MTD devices (similar to what LVM
provides for block devices)
▶ Allows to create multiple logical volumes and spread writes across all physical
blocks
▶ Takes care of managing the erase blocks and wear leveling. Makes filesystems
easier to implement
▶ Wear leveling can operate on the whole storage, not only on individual partitions
(strong advantage)
▶ Volumes can be dynamically resized or, on the opposite, can be read-only (static)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 290/520
 UBI (2)

When there is too much activity on an LEB, UBI can decide to move it to another
PEB with a lower erase count. Even read-only volumes participate to wear leveling!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 291/520
 UBI: internals

▶ UBI is storing its metadata in-band

▶ In each MTD erase block
▶ One page is reserved to count the
number of erase cycles
▶ Another page is reserved to attach
the erase block to a UBI volume
▶ The remaining pages are used to
store payload data
▶ If the device supports subpage write,
the EC and VID headers can be stored
on the same page

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 292/520
 UBI: good practice

▶ UBI is responsible for distributing writes all over the flash device: the more space
you assign to a partition attached to the UBI layer the more efficient the wear
leveling will be
▶ If you need partitioning, use UBI volumes not MTD partitions
▶ Some partitions will still have to be MTD partitions: e.g. the bootloaders and
bootloader environments
▶ If you need extra MTD partitions, try to group them at the end or the beginning
of the flash device

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 293/520
 UBI layout: bad example

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 294/520
 UBI layout: good example

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 295/520
 UBIFS

Unsorted Block Images File System

▶ http://www.linux-mtd.infradead.org/doc/ubifs.html
▶ The filesystem part of the UBI/UBIFS couple
▶ Works on top of UBI volumes
▶ Journaling file system providing better performance than JFFS2 and addressing its
scalability issues
▶ See this paper for more technical details about UBIFS internals
http://www.linux-mtd.infradead.org/doc/ubifs_whitepaper.pdf

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 296/520
 Linux: UBI host tools

▶ ubinize is the only host tool for the UBI layer

▶ Creates a UBI image to be flashed on an MTD partition
▶ Takes the following arguments:
▶ -o <output-file-path>
Path to the output image file
▶ -p <peb-size>
The PEB size (MTD erase block size)
▶ -m <min-io-size>
The minimum write unit size (e.g. MTD write size)
▶ -s <subpage-size>
Subpage size, only needed if both your flash and your flash controller are supporting
subpage writes
▶ The last argument is a path to a UBI image description file (see next page for an
example)
▶ Example: ubinize -o ubi.img -p 16KiB -m 512 -s 256 ubinize.cfg

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 297/520
 ubinize configuration file

▶ Can contain several sections

▶ Each section is describing a UBI volume
▶ static volumes are meant to store read-only blobs of data, and get the
minimum corresponding size. CRC checks are done on them.
▶ A read-only UBIFS filesystem can go in a static volume, but in this case dynamic
volumes are best for performance (CRC checking also done at UBIFS level).

[data-volume]
[rootfs-volume]
[kernel-volume] mode=ubi
mode=ubi
mode=ubi image=data.ubifs
image=rootfs.ubifs
image=zImage vol_id=3
vol_id=2
vol_id=1 vol_size=30MiB
vol_size=2MiB
vol_type=static vol_type=dynamic
vol_type=dynamic
vol_name=kernel vol_name=data
vol_name=rootfs
vol_flags=autoresize
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 298/520
 U-Boot: UBI tools
Grouped under the ubi command
▶ ubi part <part-name>
Attach an MTD partition to the UBI layer
▶ ubi info [layout]
Display UBI device information
(or volume information if the layout string is passed)
▶ ubi check <vol-name>
Check if a volume exists
▶ ubi readvol <dest-addr> <vol-name> [<size>]
Read volume contents
▶ U-Boot also provides tools to update the UBI device contents
▶ Using them is highly discouraged (the U-Boot UBI implementation is not entirely
stable, and using commands that do not touch the UBI metadata is safer)
▶ ubi createvol <vol-name> [<size>] [<type>]
▶ ubi removevol <vol-name>
▶ ubi writevol <src-addr> <vol-name> <size>
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 299/520
 Linux: UBI target tools (1)

▶ Tools used on the target to dynamically create and modify UBI elements
▶ UBI device management:
▶ ubiformat /dev/mtdx
Format an MTD partition and preserve Erase Counter information if any.
Example: ubiformat /dev/mtd1
▶ ubiattach -m <MTD-device-id> /dev/ubi_ctrl
Attach an MTD partition/device to the UBI layer, and create a UBI device
Example: ubiattach -m 1 /dev/ubi_ctrl
▶ ubidetach -m <MTD-device-id> /dev/ubi_ctrl
Detach an MTD partition/device from the UBI layer, and remove the associated UBI
device
Example: ubidetach -m 1 /dev/ubi_ctrl

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 300/520
 Linux: UBI target tools (2)

UBI volume management:

▶ ubimkvol /dev/ubi<UBI-device-id> -N <name> -s <size>
Create a new volume. Use -m in place of -s <size> if you want to assign all the
remaining space to this volume.
▶ ubirmvol /dev/ubi<UBI-device-id> -N <name>
Delete a UBI volume
▶ ubiupdatevol /dev/ubi<UBI-device-id>_<UBI-vol-id> [-s <size>] <vol-image-
file>
Update volume contents
▶ ubirsvol /dev/ubi<UBI-device-id> -N <name> -s <size>
Resize a UBI volume
▶ ubirename /dev/ubi<UBI-device-id>_<UBI-vol-id> <old-name> <new-name>
Rename a UBI volume

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 301/520
 Linux tools: BusyBox UBI limitations

Beware that the implementation of UBI commands in BusyBox is still incomplete. For
example:
▶ ubirsvol doesn’t support -N <name>. You have to use specify the volume to
resize by its id (-n num):
ubirsvol /dev/ubi0 -n 4 -s 64 MiB
▶ Same constraint for ubirmvol:
ubirmvol /dev/ubi0 -n 4

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 302/520
 Linux: UBIFS host tools

UBIFS filesystems images can be created using mkfs.ubifs

▶ mkfs.ubifs -m 4096 -e 258048 -c 1000 -r rootfs/ ubifs.img
▶ -m 4096, minimal I/O size
(see /sys/class/mtd/mtdx/writesize).
▶ -e 258048, logical erase block size (smaller than PEB size, can be found in the
kernel log after running ubiattach)
▶ -c 1000, maximum size of the UBI volume the image will be flashed into, in number
of logical erase blocks. Do not make this number unnecessary big, otherwise the
UBIFS data structures will be bigger than needed and performance will be degraded.
Details: http://linux-mtd.infradead.org/faq/ubifs.html#L_max_leb_cnt
▶ Once created
▶ Can be written to a UBI volume from the target using ubiupdatevol from Linux on
the target
▶ Or, can be included in a UBI image (using ubinize on the host)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 303/520
 Linux: UBIFS target tools

▶ No specific tools are required to create a UBIFS filesystem. An empty filesystem

is created the first time it is mounted. The same applies to JFFS2.
▶ Mounting a UBIFS filesystem is done with mount:
mount -t ubifs <ubi-device-id>:<volume-name> <mount-point>
▶ Example:
mount -t ubifs ubi0:data /data

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 304/520
 Linux: UBI image creation workflow

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 305/520
 Linux: Using a UBIFS filesystem as root filesystem

▶ You just have to pass the following information on the kernel command line:
▶ ubi.mtd=1
Attach /dev/mtd1 to the UBI layer and create ubi0
▶ rootfstype=ubifs root=ubi0:rootfs
Mount the rootfs volume on ubi0 as a UBIFS filesystem
▶ Example: rootfstype=ubifs ubi.mtd=1 root=ubi0:rootfs

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 306/520
 Summary: how to boot on a UBIFS filesystem

In U-Boot:
▶ Define partitions:
setenv mtdids ...
setenv mtdparts ...
▶ Define the base Linux kernel bootargs, specifying booting on UBIFS, the UBI
volume used as root filesystem, and the MTD partition attached to UBI. Example:
setenv bootargs_base console=ttyS0 rootfstype=ubifs root=ubi0:rootfs ubi.mtd=2 ...
▶ Define the boot command sequence, loading the U-Boot partition definitions,
loading kernel and DTB images from UBI partitions, and adding mtdparts to the
kernel command line. Example:
setenv bootcmd 'mtdparts; ubi part UBI; ubi readvol 0x81000000
kernel; ubi readvol 0x82000000 dtb; setenv bootargs ${bootargs_base}
${mtdparts}; bootz 0x81000000 - 0x82000000'

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 307/520
 Linux: Block emulation layers

▶ Sometimes we need block devices to re-use existing block filesystems, especially

read-only ones like SquashFs
▶ Linux provides two block emulation layers:
▶ mtdblock: block devices emulated on top of MTD devices
▶ ubiblock: block devices emulated on top of UBI volumes

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 308/520
 Linux: mtdblock

▶ The mtdblock layer creates a block device for each MTD device of the system
▶ Usually named /dev/mtdblockX.
▶ Allows read/write block-level access. However bad blocks are not handled, and no
wear leveling is done for writes.
▶ For historical reasons, JFFS2 and YAFFS2 filesystems require a block device for
the mount command.
▶ Do not write on mtdblock devices

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 309/520
 Linux: ubiblock

▶ CONFIG_MTD_UBI_BLOCK
▶ Implemented by Ezequiel Garcia from Bootlin.
▶ Preferred over mtdblock if UBI is available (UBI accounts for data retention and
wear leveling issues, while MTD does not)
▶ The ubiblock layer creates read-only block devices on demand
▶ The user specifies which static volumes (s)he would like to attach to ubiblock
▶ Through the kernel command line: by passing
ubi.block=<ubi-dev-id>,<volume-name>
Example: ubi.block=0,rootfs
▶ In Linux, using the ubiblock utility provided by mtd-utils:
ubiblock --create <ubi-volume-dev-file>
▶ Usually named /dev/ubiblockX_Y, where X is the UBI device id and Y is the UBI
volume id (example: /dev/ubiblock0_3)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 310/520
 Useful reading

▶ Managing flash storage with Linux:

https://bootlin.com/blog/managing-flash-storage-with-linux/
▶ Documentation on the linux-mtd website:
http://www.linux-mtd.infradead.org/
▶ Details about creating UBI and UBIFS images:
https://bootlin.com/blog/creating-flashing-ubi-ubifs-images/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 311/520
 Practical lab - Flash Filesystems

▶ Creating partitions in your internal flash storage

▶ Creating a UBI image with several volumes and
flashing it from U-Boot
▶ Manipulating UBI volumes from Linux

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 312/520
 Embedded Linux system development

Embedded Linux system

development

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 313/520
 Contents

▶ Using open-source components

▶ Tools for the target device
▶ Networking
▶ System utilities
▶ Language interpreters
▶ Audio, video and multimedia
▶ Graphical toolkits
▶ Databases
▶ Web browsers
▶ System building

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 314/520
 Embedded Linux system development

Leveraging open-source components in an

Embedded Linux system

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 315/520
 Third party libraries and applications

▶ One of the advantages of embedded Linux is the wide range of third-party libraries
and applications that one can leverage in its product
▶ They are freely available, freely distributable, and thanks to their open-source
nature, they can be analyzed and modified according to the needs of the project
▶ However, efficiently re-using these components is not always easy. One must:
▶ Find these components
▶ Choose the most appropriate ones
▶ Cross-compile them
▶ Integrate them in the embedded system and with the other applications

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 316/520
 Find existing components

▶ Free Software Directory

https://directory.fsf.org
▶ Look at other embedded Linux products, and see what their components are
▶ Look at the list of software packaged by embedded Linux build systems
▶ These are typically chosen for their suitability to embedded systems
▶ Ask the community or Google
▶ This presentation will also feature a list of components for common needs

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 317/520
 Choosing components

Not all free software components are necessarily good to re-use. One must pay
attention to:
▶ Vitality of the developer and user communities. This vitality ensures long-term
maintenance of the component, and relatively good support. It can be measured
by looking at the mailing-list traffic and the version control system activity.
▶ Quality of the component. Typically, if a component is already available through
embedded build systems, and has a dynamic user community, it probably means
that the quality is relatively good.
▶ License. The license of the component must match your licensing constraints.
For example, GPL libraries cannot be used in proprietary applications.
▶ Technical requirements. Of course, the component must match your technical
requirements. But don’t forget that you can improve the existing components if a
feature is missing!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 318/520
 Licenses (1)

▶ All software that are under a free software license give four freedoms to all users
▶ Freedom to use
▶ Freedom to study
▶ Freedom to copy
▶ Freedom to modify and distribute modified copies
▶ See https://www.gnu.org/philosophy/free-sw.html for a definition of Free
Software
▶ Open Source software, as per the definition of the Open Source Initiative, are
technically similar to Free Software in terms of freedoms
▶ See https://www.opensource.org/docs/osd for the definition of Open Source
Software

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 319/520
 Licenses (2)

▶ Free Software licenses fall in two main categories

▶ The copyleft licenses
▶ The non-copyleft licenses
▶ The concept of copyleft is to ask for reciprocity in the freedoms
given to a user.
▶ The result is that when you receive a software under a copyleft
free software license and distribute modified versions of it, you
must do so under the same license
▶ Same freedoms to the new users
▶ It’s an incentive to contribute back your changes instead of
keeping them secret
▶ Non-copyleft licenses have no such requirements, and modified
versions can be kept proprietary, but they still require attribution

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 320/520
 GPL

▶ GNU General Public License

▶ Covers around 55% of the free software projects
▶ Including the Linux kernel, Busybox and many applications
▶ Is a copyleft license
▶ Requires derivative works to be released under the same license
▶ Programs linked with a library released under the GPL must also be released under
the GPL
▶ Some programs covered by version 2 (Linux kernel, Busybox, U-Boot...)
▶ A number of programs are covered by version 3, released in 2007: gcc, bash, grub,
samba, Qt...
▶ Major change for the embedded market: the requirement that the user must be able
to run the modified versions on the device, if the device is a consumer device

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 321/520
 GPL: redistribution

▶ No obligation when the software is not distributed

▶ You can keep your modifications secret until the product delivery
▶ It is then authorized to distribute binary versions, if one of the following
conditions is met:
▶ Convey the binary with a copy of the source on a physical medium
▶ Convey the binary with a written offer valid for 3 years that indicates how to fetch
the source code
▶ Convey the binary with the network address of a location where the source code can
be found
▶ See section 6. of the GPL license
▶ In all cases, the attribution and the license must be preserved
▶ See section 4. and 5.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 322/520
 LGPL

▶ GNU Lesser General Public License

▶ Covers around 10% of the free software projects
▶ A copyleft license
▶ Modified versions must be released under the same license
▶ But, programs linked against a library under the LGPL do not need to be released
under the LGPL and can be kept proprietary.
▶ However, the user must keep the ability to update the library independently from the
program. Dynamic linking is the easiest solution. Statically linked executables are
only possible if the developer provides a way to relink with an update (with source
code or linkable object files).
▶ If this constraint is too strong for you, use a library with a more permissive license if
you can (such as the musl C library, with MIT license).
▶ Used instead of the GPL for most of the libraries, including the C libraries
▶ Some exceptions: MySQL, or Qt <= 4.4
▶ Also available in two versions, v2 and v3
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 323/520
 Licensing: examples

▶ You make modifications to the Linux kernel (to add drivers or adapt to your
board), to Busybox, U-Boot or other GPL software
▶ You must release the modified versions under the same license, and be ready to
distribute the source code to your customers
▶ You make modifications to the C library or any other LGPL library
▶ You must release the modified versions under the same license
▶ You create an application that relies on LGPL libraries
▶ You can keep your application proprietary, but you must link dynamically with the
LGPL libraries
▶ You make modifications to a non-copyleft licensed software
▶ You can keep your modifications proprietary, but you must still credit the authors

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 324/520
 Non-copyleft licenses

▶ A large family of non-copyleft licenses that are relatively similar in their

requirements
▶ A few examples
▶ Apache license (around 4%)
▶ BSD license (around 6%)
▶ MIT license (around 4%)
▶ X11 license
▶ Artistic license (around 9 %)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 325/520
 BSD license

Copyright (c) <year>, <copyright holder>

All rights reserved.

Redistribution and use in source and binary forms, with or without

modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the <organization> nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.

[...]

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 326/520
 Is this free software?

▶ Most of the free software projects are covered by 10 well-known licenses, so it is

fairly easy for the majority of projects to get a good understanding of the license
▶ Otherwise, read the license text
▶ Check Free Software Foundation’s opinion
https://www.fsf.org/licensing/licenses/
▶ Check Open Source Initiative’s opinion
https://www.opensource.org/licenses

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 327/520
 Respect free software licenses

▶ Free Software is not public domain software, the distributors have obligations due
to the licenses
▶ Before using a free software component, make sure the license matches your project
constraints
▶ Make sure to keep a complete list of the free software packages you use, the original
version numbers you used, and to keep your modifications and adaptations
well-separated from the original version.
▶ Buildroot and Yocto Project can generate this list for you!
▶ Conform to the license requirements before shipping the product to the customers.
▶ Free Software licenses have been enforced successfully in courts. Organizations
which can help:
▶ Software Freedom Law Center, https://www.softwarefreedom.org/
▶ Software Freedom Conservancy, https://sfconservancy.org/
▶ Ask your legal department!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 328/520
 Keeping changes separate (1)

▶ When integrating existing open-source components in your project, it is

sometimes needed to make modifications to them
▶ Better integration, reduced footprint, bug fixes, new features, etc.
▶ Instead of mixing these changes, it is much better to keep them separate from the
original component version
▶ If the component needs to be upgraded, easier to know what modifications were
made to the component
▶ If support from the community is requested, important to know how different the
component we’re using is from the upstream version
▶ Makes contributing the changes back to the community possible
▶ It is even better to keep the various changes made on a given component separate
▶ Easier to review and to update to newer versions

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 329/520
 Keeping changes separate (2)

▶ The simplest solution is to use Quilt

▶ Quilt is a tool that allows to maintain a stack of patches over source code
▶ Makes it easy to add, remove modifications from a patch, to add and remove
patches from stack and to update them
▶ The stack of patches can be integrated into your version control system
▶ https://savannah.nongnu.org/projects/quilt/
▶ Another solution is to use a version control system
▶ Import the original component version into your version control system
▶ Maintain your changes in a separate branch

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 330/520
 Embedded Linux system development

Tools for the target device: Networking

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 331/520
 ssh server and client: Dropbear

https://matt.ucc.asn.au/dropbear/dropbear.html
▶ Very small memory footprint ssh server for embedded systems
▶ Satisfies most needs. Both client and server!
▶ Size: 110 KB, statically compiled with uClibc on x86.
(OpenSSH client and server: approx 1200 KB, dynamically compiled with glibc on
x86)
▶ Useful to:
▶ Get a remote console on the target device
▶ Copy files to and from the target device (scp or rsync -e ssh).
▶ An alternative to OpenSSH, used on desktop and server systems.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 332/520
 Benefits of a web server interface

Many network enabled devices can just have a network interface

▶ Examples: modems / routers, IP cameras, printers...
▶ No need to develop drivers and applications for computers connected to the
device. No need to support multiple operating systems!
▶ Just need to develop static or dynamic HTML pages (possibly with powerful
client-side JavaScript).
Easy way of providing access to device information and parameters.
▶ Reduced hardware costs (no LCD, very little storage space needed)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 333/520
 Web servers

▶ BusyBox http server: https://busybox.net

▶ Tiny: only adds 9 K to BusyBox (dynamically linked with glibc
on x86, with all features enabled.)
▶ Sufficient features for many devices with a web interface,
including CGI, http authentication and script support (like PHP,
with a separate interpreter).
▶ License: GPL
▶ Other possibilities: lightweight servers like Boa, thttpd, lighttpd,
nginx, etc
▶ Some products are using Node.js, which is lightweight enough to
be used.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 334/520
 Network utilities (1)

▶ avahi is an implementation of Multicast DNS Service Discovery, that allows

programs to publish and discover services on a local network
▶ bind, a DNS server
▶ iptables, the user space tools associated to the Linux firewall, Netfilter
▶ iw and wireless tools, the user space tools associated to Wireless devices
▶ netsnmp, implementation of the SNMP protocol
▶ openntpd, implementation of the Network Time Protocol, for clock
synchronization
▶ openssl, a toolkit for SSL and TLS connections

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 335/520
 Network utilities (2)

▶ pppd, implementation of the Point to Point Protocol, used for dial-up connections
▶ samba, implements the SMB and CIFS protocols, used by Windows to share files
and printers
▶ coherence, a UPnP/DLNA implementation
▶ vsftpd, proftpd, FTP servers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 336/520
 Embedded Linux system development

Tools for the target device: System utilities

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 337/520
 System utilities

▶ dbus, an inter-application object-oriented communication bus

▶ gpsd, a daemon to interpret and share GPS data
▶ libusb, a user space library for accessing USB devices without writing an in-kernel
driver
▶ Utilities for kernel subsystems: i2c-tools for I2C, input-tools for input, mtd-utils
for MTD devices, usbutils for USB devices

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 338/520
 Embedded Linux system development

Tools for the target device: Language

interpreters

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 339/520
 Language interpreters

▶ Interpreters for the most common scripting languages are available. Useful for
▶ Application development
▶ Web services development
▶ Scripting
▶ Languages supported
▶ Lua
▶ Python
▶ Perl
▶ Ruby
▶ TCL
▶ PHP

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 340/520
 Embedded Linux system development

Tools for the target device: Audio, video and

multimedia

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 341/520
 Audio, video and multimedia

▶ GStreamer, a multimedia framework

▶ Allows to decode/encode a wide variety of codecs.
▶ Supports hardware encoders and decoders through plugins, proprietary/specific
plugins are often provided by SoC vendors.
▶ alsa-lib, the user space tools associated to the ALSA sound kernel subsystem
▶ Directly using encoding and decoding libraries, if you decide not to use GStreamer:
libavcodec, libogg, libtheora, libvpx, flac (lossless audio compression), libvorbis,
libopus (better than vorbis!), libmad, libsndfile, speex (for human speech), etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 342/520
 Embedded Linux system development

Tools for the target device: Graphical toolkits

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 343/520
 Embedded Linux system development

Graphical toolkits: “Low-level” solutions and

layers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 344/520
 X.org - KDrive

▶ Stand-alone simplified version of the X server, for embedded

systems
▶ Formerly know as Tiny-X
▶ Kdrive is integrated in the official X.org server
▶ Works on top of the Linux frame buffer, thanks to the Xfbdev
variant of the server
▶ Real X server
▶ Fully supports the X11 protocol: drawing, input event handling,
etc.
▶ Allows to use any existing X11 application or library
▶ Actively developed and maintained.
▶ X11 license
▶ https://www.x.org

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 345/520
 Kdrive: architecture

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 346/520
 Kdrive: usage

▶ Can be directly programmed using Xlib / XCB

▶ Low-level graphic library, rarely used
▶ Or, usually used with a toolkit on top of it
▶ Gtk
▶ Qt
▶ Enlightenment Foundation Libraries
▶ Others: Fltk, WxEmbedded, etc

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 347/520
 Wayland

▶ Intended to be a simpler replacement for X

▶ Wayland is a protocol for a compositor to talk to its clients as
well as a C library implementation of that protocol.
▶ Weston: a minimal and fast reference implementation of a
Wayland compositor, and is suitable for many embedded and
mobile use cases.
▶ Not fully deployed yet. However, the ports of Gtk and Qt to
Wayland are complete.
▶ https://wayland.freedesktop.org/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 348/520
 Wayland: architecture

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 349/520
 Embedded Linux system development

Graphical toolkits: “High-level” solutions

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 350/520
 Gtk

▶ The famous toolkit, providing widget-based high-level APIs to

develop graphical applications
▶ Standard API in C, but bindings exist for various languages:
C++, Python, etc.
▶ Works on top of X.org and Wayland.
▶ No windowing system, a lightweight window manager needed to
run several applications. Possible solution: Matchbox.
▶ License: LGPL
▶ Multiplatform: Linux, MacOS, Windows.
▶ https://www.gtk.org

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 351/520
 Gtk stack components

▶ Glib, core infrastructure

▶ Object-oriented infrastructure GObject
▶ Event loop, threads, asynchronous queues, plug-ins, memory allocation, I/O
channels, string utilities, timers, date and time, internationalization, simple XML
parser, regular expressions
▶ Data types: memory slices and chunks, linked lists, arrays, trees, hash tables, etc.
▶ Pango, internationalization of text handling
▶ ATK, accessibility toolkit
▶ Cairo, vector graphics library
▶ Gtk+, the widget library itself
▶ The Gtk stack is a complete framework to develop applications

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 352/520
 Gtk example

Maemo tablet / phone interface

GTK is losing traction, however: Mer, the descendent of Maemo, is now implemented
in EFL (see next slides).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 353/520
 Qt (1)

▶ The other famous toolkit, providing widget-based high-level APIs to develop

graphical applications
▶ Implemented in C++
▶ the C++ library is required on the target system
▶ standard API in C++, but with bindings for other languages
▶ Works either on top of
▶ Framebuffer
▶ X11
▶ Wayland
▶ Multiplatform: Linux, MacOS, Windows.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 354/520
 Qt (2)

▶ Qt is more than just a graphical toolkit, it also offers a complete development

framework: data structures, threads, network, databases, XML, etc.
▶ See our presentation Qt for non graphical applications presentation at ELCE 2011
(Thomas Petazzoni): https://j.mp/W4PK85
▶ Qt Embedded has an integrated windowing system, allowing several applications
to share the same screen
▶ Very well documented
▶ Since version 4.5, available under the LGPL, allowing proprietary applications
▶ Since version 4.7, LGPL v2 is dropped, leaving LGPLv3 and GPLv3 (and GPLv2
for some parts) as only options, which add more restrictions to proprietary
applications.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 355/520
 Qt’s usage

Source: https://www.qt.io/qt-for-device-creation/
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 356/520
 Other less frequent solutions

▶ Enlightenment Foundation Libraries (EFL)

▶ Very powerful. Supported by Samsung, Intel and Free.fr.
▶ Work on top of X or Wayland.
▶ https://www.enlightenment.org/about-efl

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 357/520
 Embedded Linux system development

Tools for the target device: Databases

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 358/520
 Lightweight database - SQLite

https://www.sqlite.org
▶ SQLite is a small C library that implements a self-contained, embeddable,
lightweight, zero-configuration SQL database engine
▶ The database engine of choice for embedded Linux systems
▶ Can be used as a normal library
▶ Can be directly embedded into a application, even a proprietary one since SQLite is
released in the public domain

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 359/520
 Embedded Linux system development

Tools for the target device: Web browsers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 360/520
 WebKit

https://webkit.org/
▶ Web browser engine. Application framework that can be used to
develop web browsers.
▶ License: portions in LGPL and others in BSD. Proprietary
applications allowed.
▶ Used by many web browsers: Safari, iPhone and Android default
browsers ... Google Chrome now uses a fork of its WebCore
component). Used by e-mail clients too to render HTML.
▶ Multiple graphical back-ends: Qt4, GTK, EFL...
▶ You could use it to create your custom browser.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 361/520
 Embedded Linux system development

System building

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 362/520
 System building: goal and solutions

▶ Goal
▶ Integrate all the software components, both
third-party and in-house, into a working root
filesystem
▶ It involves the download, extraction,
configuration, compilation and installation of
all components, and possibly fixing issues and
adapting configuration files
▶ Several solutions
▶ Manually
▶ System building tools
▶ Distributions or ready-made filesystems
Penguin picture: https://bit.ly/1PwDklz

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 363/520
 System building: manually

▶ Manually building a target system involves downloading, configuring, compiling

and installing all the components of the system.
▶ All the libraries and dependencies must be configured, compiled and installed in
the right order.
▶ Sometimes, the build system used by libraries or applications is not very
cross-compile friendly, so some adaptations are necessary.
▶ There is no infrastructure to reproduce the build from scratch, which might cause
problems if one component needs to be changed, if somebody else takes over the
project, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 364/520
 System building: manually (2)

▶ Manual system building is not recommended for production projects

▶ However, using automated tools often requires the developer to dig into specific
issues
▶ Having a basic understanding of how a system can be built manually is therefore
very useful to fix issues encountered with automated tools
▶ We will first study manual system building, and during a practical lab, create a
system using this method
▶ Then, we will study the automated tools available, and use one of them during a lab

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 365/520
 System foundations

▶ A basic root file system needs at least

▶ A traditional directory hierarchy, with /bin, /etc, /lib, /root, /usr/bin,
/usr/lib, /usr/share, /usr/sbin, /var, /sbin
▶ A set of basic utilities, providing at least the init program, a shell and other
traditional UNIX command line tools. This is usually provided by Busybox
▶ The C library and the related libraries (thread, math, etc.) installed in /lib
▶ A few configuration files, such as /etc/inittab, and initialization scripts in
/etc/init.d
▶ On top of this foundation common to most embedded Linux system, we can add
third-party or in-house components

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 366/520
 Target and build spaces

▶ The system foundation, Busybox and C library, are the core of the target root
filesystem
▶ However, when building other components, one must distinguish two directories
▶ The target space, which contains the target root filesystem, everything that is
needed for execution of the application
▶ The build space, which will contain a lot more files than the target space, since it is
used to keep everything needed to compile libraries and applications. So we must
keep the headers, documentation, and other configuration files

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 367/520
 Build systems

Each open-source component comes with a mechanism to configure, compile and

install it
▶ A basic Makefile
▶ Need to read the Makefile to understand how it works and how to tweak it for
cross-compilation
▶ A build system based on the Autotools
▶ As this is the most common build system, we will study it in details
▶ CMake, https://cmake.org/
▶ Newer and simpler than the autotools. Used by (sometimes large) projects such as
KDE, KiCad, LLVM / Clang, Scribus, OpenCV. Used by Netflix too.
▶ Scons, https://www.scons.org/
▶ Waf, https://github.com/waf-project/waf
▶ Other manual build systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 368/520
 Autotools and friends

▶ A family of tools, which associated together form a complete and extensible build
system
▶ autoconf is used to handle the configuration of the software package
▶ automake is used to generate the Makefiles needed to build the software package
▶ pkgconfig is used to ease compilation against already installed shared libraries
▶ libtool is used to handle the generation of shared libraries in a system-independent
way
▶ Most of these tools are old and relatively complicated to use, but they are used by
a majority of free software packages today. One must have a basic understanding
of what they do and how they work.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 369/520
 automake / autoconf / autoheader

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 370/520
 automake / autoconf

▶ Files written by the developer

▶ configure.in describes the configuration options and the checks done at configure
time
▶ Makefile.am describes how the software should be built
▶ The configure script and the Makefile.in files are generated by autoconf and
automake respectively.
▶ They should never be modified directly
▶ They are usually shipped pre-generated in the software package, because there are
several versions of autoconf and automake, and they are not completely compatible
▶ The Makefile files are generated at configure time, before compiling
▶ They are never shipped in the software package.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 371/520
 Configuring and compiling: native case

▶ The traditional steps to configure and compile an autotools based package are
▶ Configuration of the package
./configure
▶ Compilation of the package
make
▶ Installation of the package
make install
▶ Additional arguments can be passed to the ./configure script to adjust the
component configuration (run ./configure --help)
▶ Only the make install target needs to be done as root if the installation should
take place system-wide

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 372/520
 Configuring and compiling: cross case (1)

▶ For cross-compilation, things are a little bit more complicated.

▶ At least some of the environment variables AR, AS, LD, NM, CC, GCC, CPP, CXX,
STRIP, OBJCOPY must be defined to point to the proper cross-compilation tools.
The host tuple is also by default used as prefix.
▶ configure script arguments:
▶ --host: mandatory but a bit confusing. Corresponds to the target platform the
code will run on. Example: --host=arm-linux
▶ --build: build system. Automatically detected.
▶ --target is only for tools generating code.
▶ It is recommended to pass the --prefix argument. It defines from which location
the software will run in the target environment. We recommend /usr instead of
the default setting (/usr/local).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 373/520
 Configuring and compiling: cross case (2)

▶ If one simply runs make install, the software will be installed in the directory
passed as --prefix. For cross-compiling, one must pass the DESTDIR argument
to specify where the software must be installed.
▶ Making the distinction between the prefix (as passed with --prefix at configure
time) and the destination directory (as passed with DESTDIR at installation time)
is very important.
▶ Example:

export PATH=/usr/local/arm-linux/bin:$PATH
export CC=arm-linux-gcc
export STRIP=arm-linux-strip
./configure --host=arm-linux --prefix=/usr
make
make DESTDIR=$HOME/work/rootfs install

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 374/520
 Installation (1)

▶ The autotools based software packages provide both a install and

install-strip make targets, used to install the software, either stripped or
unstripped.
▶ For applications, the software is usually installed in <prefix>/bin, with
configuration files in <prefix>/etc and data in
<prefix>/share/<application>/
▶ The case of libraries is a little more complicated:
▶ In <prefix>/lib, the library itself (a .so.<version>), a few symbolic links, and the
libtool description file (a .la file)
▶ The pkgconfig description file in <prefix>/lib/pkgconfig
▶ Include files in <prefix>/include/
▶ Sometimes a <libname>-config program in <prefix>/bin
▶ Documentation in <prefix>/share/man or <prefix>/share/doc/

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 375/520
 Installation (2)
Contents of usr/lib after installation of libpng and zlib
▶ libpng libtool description files
./lib/libpng12.la
./lib/libpng.la -> libpng12.la
▶ libpng static version
./lib/libpng12.a
./lib/libpng.a -> libpng12.a
▶ libpng dynamic version
./lib/libpng.so.3.32.0
./lib/libpng12.so.0.32.0
./lib/libpng12.so.0 -> libpng12.so.0.32.0
./lib/libpng12.so -> libpng12.so.0.32.0
./lib/libpng.so -> libpng12.so
./lib/libpng.so.3 -> libpng.so.3.32.0
▶ libpng pkg-config description files
./lib/pkgconfig/libpng12.pc
./lib/pkgconfig/libpng.pc -> libpng12.pc
▶ zlib dynamic version
./lib/libz.so.1.2.3
./lib/libz.so -> libz.so.1.2.3
./lib/libz.so.1 -> libz.so.1.2.3
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 376/520
 Installation in the build and target spaces

▶ From all these files, everything except documentation is necessary to build an

application that relies on libpng.
▶ These files will go into the build space
▶ However, only the library .so binaries in <prefix>/lib and some symbolic links
are needed to execute the application on the target.
▶ Only these files will go in the target space
▶ The build space must be kept in order to build other applications or recompile
existing applications.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 377/520
 pkg-config

▶ pkg-config is a tool that allows to query a small database to get information on

how to compile programs that depend on libraries
▶ The database is made of .pc files, installed by default in
<prefix>/lib/pkgconfig/.
▶ pkg-config is used by the configure script to get the library configurations
▶ It can also be used manually to compile an application:
arm-linux-gcc -o test test.c $(pkg-config --libs --cflags thelib)
▶ By default, pkg-config looks in /usr/lib/pkgconfig for the *.pc files, and
assumes that the paths in these files are correct.
▶ PKG_CONFIG_LIBDIR allows to set another location for the *.pc files and
PKG_CONFIG_SYSROOT_DIR to prepend a prefix to the paths mentioned in the .pc
files.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 378/520
 Let’s find the libraries

▶ When compiling an application or a library that relies on other libraries, the build
process by default looks in /usr/lib for libraries and /usr/include for headers.
▶ The first thing to do is to set the CFLAGS and LDFLAGS environment variables:
export CFLAGS=-I/my/build/space/usr/include/
export LDFLAGS=-L/my/build/space/usr/lib
▶ The libtool files (.la files) must be modified because they include the absolute
paths of the libraries:
- libdir='/usr/lib'
+ libdir='/my/build/space/usr/lib’
▶ The PKG_CONFIG_LIBDIR environment variable must be set to the location of the
.pc files and the PKG_CONFIG_SYSROOT_DIR variable must be set to the build space
directory.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 379/520
 Further details about autotools

See our ”Demystification tutorial” about the GNU Autotools (Thomas Petazzoni,
2016)
▶ Presentation slides: https://bit.ly/2hjpojv
▶ Video: https://youtu.be/_zX8LJ9Xjyk

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 380/520
 Practical lab - Third party libraries and applications

▶ Manually cross-compiling applications and

libraries
▶ Learning about common techniques and issues.
▶ Compile and run an audio player application!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 381/520
 System building tools: principle

▶ Different tools are available to automate the process of building a target system,
including the kernel, and sometimes the toolchain.
▶ They automatically download, configure, compile and install all the components in
the right order, sometimes after applying patches to fix cross-compiling issues.
▶ They already contain a large number of packages, that should fit your main
requirements, and are easily extensible.
▶ The build becomes reproducible, which allows to easily change the configuration
of some components, upgrade them, fix bugs, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 382/520
 Available system building tools

Large choice of tools

▶ Buildroot, developed by the community
https://buildroot.org
See our dedicated course and training materials:
https://bootlin.com/training/buildroot/
▶ PTXdist, developed by Pengutronix
https://www.ptxdist.org
▶ OpenWRT, originally a fork of Buildroot for wireless routers, now a more generic project
https://openwrt.org
▶ OpenEmbedded, more flexible but also far more complicated
http://www.openembedded.org, its industrialized version Yocto Project and
vendor-specific derivatives such as Arago.
See our dedicated course and training materials:
https://bootlin.com/training/yocto/.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 383/520
 Buildroot (1)

▶ Allows to build a toolchain, a root filesystem image with many applications and
libraries, a bootloader and a kernel image
▶ Or any combination of the previous items
▶ Supports building uClibc, glibc and musl toolchains, either built by Buildroot, or
external
▶ Over 2000+ applications or libraries integrated, from basic utilities to more
elaborate software stacks: X.org, GStreamer, Qt, Gtk, WebKit, Python, PHP, etc.
▶ Good for small to medium embedded systems, with a fixed set of features
▶ No support for generating packages (.deb or .ipk)
▶ Needs complete rebuild for most configuration changes.
▶ Active community, releases published every 3 months.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 384/520
 Buildroot (2)

▶ Configuration takes place through a *config

interface similar to the kernel
make menuconfig
▶ Allows to define
▶ Architecture and specific CPU
▶ Toolchain configuration
▶ Set of applications and libraries to integrate
▶ Filesystem images to generate
▶ Kernel and bootloader configuration
▶ Build by just running
make

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 385/520
 Buildroot: adding a new package (1)

▶ A package allows to integrate a user application or library to Buildroot

▶ Each package has its own directory (such as package/gqview). This directory
contains:
▶ A Config.in file (mandatory), describing the configuration options for the package.
At least one is needed to enable the package. This file must be sourced from
package/Config.in
▶ A gqview.mk file (mandatory), describing how the package is built.
▶ A .hash file (optional, but recommended), containing hashes for the files to
download.
▶ Patches (optional). Each file of the form *.patch will be applied as a patch.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 386/520
 Buildroot: adding a new package (2)

▶ For a simple package with a single configuration option to enable/disable it, the
Config.in file looks like:
config BR2_PACKAGE_GQVIEW
bool "gqview"
depends on BR2_PACKAGE_LIBGTK2
help
GQview is an image viewer for UNIX operating systems

http://prdownloads.sourceforge.net/gqview
▶ It must be sourced from package/Config.in:

source "package/gqview/Config.in"

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 387/520
 Buildroot: adding new package (3)

▶ Create the gqview.mk file to describe the build steps

GQVIEW_VERSION = 2.1.5
GQVIEW_SOURCE = gqview-$(GQVIEW_VERSION).tar.gz
GQVIEW_SITE = http://prdownloads.sourceforge.net/gqview
GQVIEW_DEPENDENCIES = host-pkgconf libgtk2
GQVIEW_CONF_ENV = LIBS="-lm"
GQVIEW_LICENSE = GPL-2.0
GQVIEW_LICENSE_FILES = COPYING

$(eval $(autotools-package))

▶ The package directory and the prefix of all variables must be identical to the suffix
of the main configuration option BR2_PACKAGE_GQVIEW
▶ The autotools-package infrastructure knows how to build autotools packages.
A more generic generic-package infrastructure is available for packages not
using the autotools as their build system.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 388/520
 OpenEmbedded / Yocto Project

▶ The most versatile and powerful embedded Linux build system

▶ A collection of recipes (.bb files)
▶ A tool that processes the recipes: bitbake
▶ Integrates 2000+ application and libraries, is highly configurable, can generate
binary packages to make the system customizable, supports multiple
versions/variants of the same package, no need for full rebuild when the
configuration is changed.
▶ Configuration takes place by editing various configuration files
▶ Good for larger embedded Linux systems, or people looking for more
configurability and extensibility
▶ Drawbacks: very steep learning curve, very long first build.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 389/520
 Distributions - Debian

Debian GNU/Linux, https://www.debian.org

▶ Provides the easiest environment for quickly building prototypes and
developing applications. Countless runtime and development packages
available.
▶ But probably too costly to maintain and unnecessarily big for
production systems.
▶ Available on multiple architectures: ARM (armel, armhf, arm64),
MIPS, PowerPC, RiscV (in progress)...
▶ Software is compiled natively by default.
▶ Use the debootstrap command to build a root filesystem for your
architecture, with a custom selection of packages.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 390/520
 Distributions - Others

Fedora
▶ https:
//fedoraproject.org/wiki/Architectures/ARM
Image credits: https://bit.ly/2EzmJLF
▶ Supported on various recent ARM boards (such as
Beaglebone Black and Raspberry Pi)
▶ Supports QEMU emulated ARM boards too (Versatile
Express board)
▶ Shipping the same version as for desktops!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 391/520
 Embedded distributions

Even if you don’t use them for final products, they can be useful to make demos quickly

▶ Android: https://www.android.com/
Google’s distribution for phones, tablets, TVs, cars...
Except the Linux kernel, very different user space than other Linux
distributions. Very successful, lots of applications available (many
proprietary).
▶ Alpine Linux: https://www.alpinelinux.org/
Security oriented distribution based on Musl and BusyBox, fitting in
about 130 MB of storage, supporting x86 and arm, both 32 and 64 bit.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 392/520
 Application frameworks

Not real distributions you can download. Instead, they implement

middleware running on top of the Linux kernel and allowing to develop
applications.
▶ Tizen: https://www.tizen.org/
Targeting smartphones, wearables (wathes), smart TVs and In Vehicle
Infotainment devices.
Supported by big phone manufacturers (mostly Samsung) and
operators
HTML5 base application framework.
Wikipedia: 21% of the smart TVs market share in 2018
See https://en.wikipedia.org/wiki/Tizen

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 393/520
 Practical lab - Buildroot

▶ Rebuild the same system, this time with

Buildroot.
▶ See how easier it gets!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 394/520
 Embedded Linux application development

Embedded Linux
application development

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 395/520
 Contents

▶ Application development
▶ Developing applications on embedded Linux
▶ Building your applications
▶ Source management
▶ Integrated development environments (IDEs)
▶ Version control systems
▶ Debugging and analysis tools
▶ Debuggers
▶ Memory checkers
▶ System analysis

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 396/520
 Embedded Linux application development

Developing applications on embedded Linux

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 397/520
 Application development

▶ An embedded Linux system is just a normal Linux system, with usually a smaller
selection of components
▶ In terms of application development, developing on embedded Linux is exactly the
same as developing on a desktop Linux system
▶ All existing skills can be re-used, without any particular adaptation
▶ All existing libraries, either third-party or in-house, can be integrated into the
embedded Linux system
▶ Taking into account, of course, the limitation of the embedded systems in terms of
performance, storage and memory
▶ Application development could start on x86, even before the hardware is available.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 398/520
 Programming language

▶ The default programming language for system-level applications in Linux is usually

C
▶ The C library is already present on your system, nothing to add
▶ C++ can be used for larger applications
▶ The C++ library must be added to the system
▶ Some libraries, including Qt, are developed in C++ so they need the C++ library on
the system anyway
▶ Scripting languages can also be useful for quick application development, web
applications or scripts
▶ But they require an interpreter on the embedded system and have usually higher
memory consumption and slightly lower performance
▶ Most popular: Python, shell
▶ All programming languages can be used: Lua, Ada, Java, Rust...

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 399/520
 C library or higher-level libraries?

▶ For many applications, the C library already provides a relatively large set of
features
▶ file and device I/O, networking, threads and synchronization, inter-process
communication
▶ Thoroughly described in the glibc manual, or in any Linux system programming book
▶ However, the API carries a lot of history and is not necessarily easy to grasp for new
comers
▶ Therefore, using a higher level framework, such as Qt or the Gtk/Glib stack,
might be a good idea
▶ These frameworks are not only graphical libraries, their core is separate from the
graphical part
▶ But of course, these libraries have some memory and storage footprint, in the order
of a few megabytes

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 400/520
 Building your applications

▶ For simple applications that do not need to be really portable or provide

compile-time configuration options, a simple Makefile will be sufficient
▶ For more complicated applications, or if you want to be able to run your
application on a desktop Linux PC and on the target device, using a build system
is recommended
▶ autotools is ancient, complicated but very widely used.
▶ We recommend to invest in CMake instead: modern, simpler, smaller but growing
user base.
▶ The QT library is a special case, since it comes with its own build system for
applications, called qmake.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 401/520
 Simple Makefile (1)

Case of an application that only uses the C library, contains two source files and
generates a single binary
CROSS_COMPILE?=arm-linux-
CC=$(CROSS_COMPILE)gcc
OBJS=foo.o bar.o

all: foobar

foobar: $(OBJS)
Tab $(CC) -o $@ $^

clean:
Tab $(RM) -f foobar $(OBJS)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 402/520
 Simple Makefile (2)

Case of an application that uses the Glib and the GPS libraries
CROSS_COMPILE?=arm-linux-
LIBS=libgps glib-2.0
OBJS=foo.o bar.o

CC=$(CROSS_COMPILE)gcc
CFLAGS=$(shell pkg-config --cflags $(LIBS))
LDFLAGS=$(shell pkg-config --libs $(LIBS))

all: foobar

foobar: $(OBJS)
Tab $(CC) -o $@ $^ $(LDFLAGS)

clean:
Tab $(RM) -f foobar $(OBJS)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 403/520
 Embedded Linux application development

Integrated Development Environments (IDE)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 404/520
 Visual Studio Code

https://code.visualstudio.com/
▶ Created by Microsoft
▶ License: MIT
▶ Extensible, language agnostic text editor
▶ Built-in git commands
▶ The most popular IDE (open-source and proprietary)
according Stack Overflow’s 2019 survey
▶ Try it on Ubuntu: sudo snap install --classic code
▶ https://en.wikipedia.org/wiki/Visual_Studio_Code

Image credits (Wikipedia):

https://frama.link/RjFqWGBS

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 405/520
 Eclipse (1)

https://www.eclipse.org/
▶ An extensible, plug-in based software development kit, typically
used for creating IDEs.
▶ Supported by the Eclipse foundation, a non-profit consortium of image credits:
https://bit.ly/2Hntsvf
major software industry vendors (IBM, Intel, Borland, Nokia,
Wind River, Zend, Computer Associates...).
▶ Free Software license (Eclipse Public License). Incompatible with
the GPL.
▶ Supported platforms: GNU/Linux, UNIX, Windows
Extremely popular: created a lot of attraction.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 406/520
 Eclipse (2)

▶ Eclipse is actually a platform composed of many projects:

https://www.eclipse.org/projects/
▶ Some projects are dedicated to integrating into Eclipse features useful for embedded
developers (cross-compilation, remote development, remote debugging, etc.)
▶ The platform is used by major embedded Linux software vendors for their
(proprietary) system development kits: MontaVista DevRocket, TimeSys
TimeStorm, Wind River Workbench, Sysgo ELinOS.
▶ Used by free software build systems and development environments too, such as
Yocto Project and Buildroot.
Eclipse is a huge project. It would require an entire training session!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 407/520
 Other popular solutions
Vim
▶ Many embedded Linux developers simply use
Vim or Emacs. They can integrate with
debuggers, source code browsers such as
cscope, offer syntax highlighting and more.
▶ People also use QtCreator, even for non Qt
projects
▶ Atom is a very popular text editor too
▶ See Stack Overflow’s survey of most popular Emacs
IDEs (2019): https://frama.link/bfPgbb88
All these tools are available in most Linux
distributions, simply install them and try them out!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 408/520
 Embedded Linux application development

Version control systems

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 409/520
 Version control systems

Real projects can’t do without them

▶ Allow multiple developers to contribute on the same project. Each developer can
see the latest changes from the others, or choose to stick with older versions of
some components.
▶ Allow to keep track of changes, and revert them if needed.
▶ Allow developers to have their own development branch (branching)
▶ Supposed to help developers resolving conflicts with different branches (merging)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 410/520
 Traditional version control systems

Rely on a central repository. Subversion is the last popular open-source one:

▶ https://subversion.apache.org/
▶ Created as a replacement of the old CVS, removing many of its limitations.
▶ Commits on several files, proper renaming support, better performance, etc.
▶ The user interface is very similar to CVS
▶ https://en.wikipedia.org/wiki/Subversion_(software)
No longer recommended for new projects. Use distributed source control systems
instead!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 411/520
 Distributed source control systems (1)

No longer have a central repository

▶ More adapted to the way the Free Software community develops software and
organizes
▶ Allows each developer to have a full local history of the project, to create local
branches. Makes each developer’s work easier.
▶ People get working copies from other people’s working copies, and exchange
changes between themselves. Branching and merging is made easier.
▶ Make it easier for new developers to join, making their own experiments without
having to apply for repository access.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 412/520
 Distributed source control systems (2)

▶ Git
▶ Initially designed and developed by Linus Torvalds for Linux kernel development
▶ Extremely popular in the community, and used by more and more projects (kernel,
U-Boot, Barebox, uClibc, GNOME, X.org, etc.)
▶ Outstanding performance, in particular in big projects
▶ https://en.wikipedia.org/wiki/Git_(software)
▶ Mercurial
▶ Another system, created with the same goals as Git.
▶ Used by some big projects too
▶ https://en.wikipedia.org/wiki/Mercurial

https://en.wikipedia.org/wiki/Version_control_systems#Distributed_
revision_control

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 413/520
 Embedded Linux application development

Debuggers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 414/520
 GDB

The GNU Project Debugger

https://www.gnu.org/software/gdb/
▶ The debugger on GNU/Linux, available for most embedded
architectures.
▶ Supported languages: C, C++, Pascal, Objective-C, Fortran,
Ada...
▶ Console interface (useful for remote debugging).
▶ Graphical front-ends available.
▶ Can be used to control the execution of a program, set
breakpoints or change internal variables. You can also use it to
see what a program was doing when it crashed (by loading its
memory image, dumped into a core file).
See also https://en.wikipedia.org/wiki/Gdb

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 415/520
 GDB crash course
A few useful GDB commands
▶ break foobar
puts a breakpoint at the entry of function foobar()
▶ break foobar.c:42
puts a breakpoint in foobar.c, line 42
▶ print var or print task->files[0].fd
prints the variable var, or a more complicated reference. GDB can also nicely
display structures with all their members
▶ continue
continue the execution
▶ next
continue to the next line, stepping over function calls
▶ step
continue to the next line, entering into subfunctions
▶ backtrace
display the program stack
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 416/520
 GDB graphical front-ends

▶ DDD - Data Display Debugger

https://www.gnu.org/software/ddd/
A popular graphical front-end, with advanced data plotting capabilities.
▶ GDB/Insight
https://sourceware.org/insight/
From the GDB maintainers.
▶ KDbg
http://www.kdbg.org/
Another front-end, for the K Display Environment.
▶ Integration with other IDEs: Eclipse, Emacs, KDevelop, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 417/520
 Embedded Linux application development

Remote debugging

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 418/520
 Remote debugging

▶ In a non-embedded environment, debugging takes place using gdb or one of its

front-ends.
▶ gdb has direct access to the binary and libraries compiled with debugging symbols.
▶ However, in an embedded context, the target platform environment is often too
limited to allow direct debugging with gdb (2.4 MB on x86).
▶ Remote debugging is preferred
▶ ARCH-linux-gdb is used on the development workstation, offering all its features.
▶ gdbserver is used on the target system (only 100 KB on arm).

ARCH-linux-gdb
gdbserver

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 419/520
 Remote debugging: architecture

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 420/520
 Remote debugging: usage

▶ On the target, run a program through gdbserver.

Program execution will not start immediately.
gdbserver localhost:<port> <executable> <args>
gdbserver /dev/ttyS0 <executable> <args>
▶ Otherwise, attach gdbserver to an already running program:
gdbserver --attach localhost:<port> <pid>
▶ Then, on the host, start ARCH-linux-gdb <executable>,
and use the following gdb commands:
▶ To connect to the target:
gdb> target remote <ip-addr>:<port> (networking)
gdb> target remote /dev/ttyS0 (serial link)
▶ To tell gdb where shared libraries are:
gdb> set sysroot <library-path> (without lib/)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 421/520
 Post mortem analysis

▶ When an application crashes due to a segmentation fault and the application was
not under control of a debugger, we get no information about the crash
▶ Fortunately, Linux can generate a core file that contains the image of the
application memory at the moment of the crash, and gdb can use this core file to
let us analyze the state of the crashed application
▶ On the target
▶ Use ulimit -c unlimited to enable the generation of a core file when a crash
occurs
▶ On the host
▶ After the crash, transfer the core file from the target to the host, and run
ARCH-linux-gdb -c core-file application-binary

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 422/520
 Embedded Linux application development

Memory checkers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 423/520
 Valgrind (1)

https://valgrind.org/
▶ GNU GPL Software suite for debugging and profiling programs.
▶ Supported platforms: Linux on x86, x86_64, arm (armv7 only),
arm64, mips32, s390, ppc32 and ppc64. Also supported on other
operating systems (Android, Darwin, Illumos, Solaris...)
▶ Can detect many memory management and threading bugs.
▶ Profiler: provides information helpful to speed up your program
and reduce its memory usage.
▶ The most popular tool for this usage. Even used by projects with
hundreds of programmers.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 424/520
 Valgrind (2)

▶ Can be used to run any program, without the need to recompile

it.
▶ Example usage
valgrind --leak-check=yes ls -la
▶ Works by adding its own instrumentation to your code and then
running in on its own virtual cpu core.
Significantly slows down execution, but still fine for testing!
▶ More details on https://valgrind.org/info/ and
https://valgrind.org/docs/manual/coregrind_core.html

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 425/520
 Embedded Linux application development

System analysis

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 426/520
 strace
System call tracer
https://sourceforge.net/projects/strace/
▶ Available on all GNU/Linux systems
Can be built by your cross-compiling toolchain generator.
▶ Even easier: drop a ready-made static binary for your architecture, just when you
need it. See
https://github.com/bootlin/static-binaries/tree/master/strace
▶ Allows to see what any of your processes is doing:
accessing files, allocating memory...
Often sufficient to find simple bugs.
▶ Usage:
strace <command> (starting a new process)
strace -p <pid> (tracing an existing process) strace -c <command> (statistics
of system calls taking most time)
See man strace for details.
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 427/520
 strace example output

> strace cat Makefile

execve("/bin/cat", ["cat", "Makefile"], [/* 38 vars */]) = 0
brk(0) = 0x98b4000
access("/etc/ld.so.nohwcap", F_OK) = -1 ENOENT (No such file or directory)
mmap2(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7f85000
access("/etc/ld.so.preload", R_OK) = -1 ENOENT (No such file or directory)
open("/etc/ld.so.cache", O_RDONLY) = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=111585, ...}) = 0
mmap2(NULL, 111585, PROT_READ, MAP_PRIVATE, 3, 0) = 0xb7f69000
close(3) = 0
access("/etc/ld.so.nohwcap", F_OK) = -1 ENOENT (No such file or directory)
open("/lib/tls/i686/cmov/libc.so.6", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\320h\1\0004\0\0\0\344"..., 512) = 512
fstat64(3, {st_mode=S_IFREG|0755, st_size=1442180, ...}) = 0
mmap2(NULL, 1451632, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0xb7e06000
mprotect(0xb7f62000, 4096, PROT_NONE) = 0
mmap2(0xb7f63000, 12288, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x15c) = 0xb7f63000
mmap2(0xb7f66000, 9840, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0xb7f66000
close(3) = 0

Hint: follow the open file descriptors returned by open().

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 428/520
 strace -c example output

> strace -c cheese

% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
36.24 0.523807 19 27017 poll
28.63 0.413833 5 75287 115 ioctl
25.83 0.373267 6 63092 57321 recvmsg
3.03 0.043807 8 5527 writev
2.69 0.038865 10 3712 read
2.14 0.030927 3 10807 getpid
0.28 0.003977 1 3341 34 futex
0.21 0.002991 3 1030 269 openat
0.20 0.002889 2 1619 975 stat
0.18 0.002534 4 568 mmap
0.13 0.001851 5 356 mprotect
0.10 0.001512 2 784 close
0.08 0.001171 3 461 315 access
0.07 0.001036 2 538 fstat
...

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 429/520
 ltrace

A tool to trace library calls used by a program and all the signals it receives
▶ Very useful complement to strace, which shows only system calls.
▶ Of course, works even if you don’t have the sources
▶ Allows to filter library calls with regular expressions, or just by a list of function
names.
▶ Manual page: https://linux.die.net/man/1/ltrace
See https://en.wikipedia.org/wiki/Ltrace for details

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 430/520
 ltrace example output

ltrace nedit index.html

sscanf(0x8274af1, 0x8132618, 0x8248640, 0xbfaadfe8, 0) = 1
sprintf("const 0", "const %d", 0) = 7
strcmp("startScan", "const 0") = 1
strcmp("ScanDistance", "const 0") = -1
strcmp("const 200", "const 0") = 1
strcmp("$list_dialog_button", "const 0") = -1
strcmp("$shell_cmd_status", "const 0") = -1
strcmp("$read_status", "const 0") = -1
strcmp("$search_end", "const 0") = -1
strcmp("$string_dialog_button", "const 0") = -1
strcmp("$rangeset_list", "const 0") = -1
strcmp("$calltip_ID", "const 0") = -1

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 431/520
 ltrace summary
Example summary at the end of the ltrace output (-c option)
Process 17019 detached
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
100.00 0.000050 50 1 set_thread_area
0.00 0.000000 0 48 read
0.00 0.000000 0 44 write
0.00 0.000000 0 80 63 open
0.00 0.000000 0 19 close
0.00 0.000000 0 1 execve
0.00 0.000000 0 2 2 access
0.00 0.000000 0 3 brk
0.00 0.000000 0 1 munmap
0.00 0.000000 0 1 uname
0.00 0.000000 0 1 mprotect
0.00 0.000000 0 19 mmap2
0.00 0.000000 0 50 46 stat64
0.00 0.000000 0 18 fstat64
------ ----------- ----------- --------- --------- ----------------
100.00 0.000050 288 111 total

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 432/520
 OProfile

http://oprofile.sourceforge.net
▶ A system-wide profiling tool
▶ Can collect statistics like the top users of the CPU.
▶ Works without having the sources.
▶ Requires a kernel patch to access all features, but is already available in a
standard kernel.
▶ Requires more investigation to see how it works.
▶ Ubuntu/Debian packages: oprofile, oprofile-gui

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 433/520
 Practical lab - App. development and debugging

Application development
▶ Compile your own application with the ncurses
library
Remote debugging
▶ Set up remote debugging tools on the target:
strace, ltrace
and gdbserver.
▶ Debug a simple application running on the
target using remote debugging

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 434/520
 Real-time in embedded Linux systems

Real-time in embedded
Linux systems

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 435/520
 Real-time in embedded Linux systems

Introduction

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 436/520
 Embedded Linux and real time

▶ Due to its advantages, Linux and open-source software are more and more
commonly used in embedded applications
▶ However, some applications also have real-time constraints
▶ They, at the same time, want to
▶ Get all the nice advantages of Linux: hardware support, components re-use, low
cost, etc.
▶ Get their real-time constraints met

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 437/520
 Embedded Linux and real time

▶ Linux is an operating system part of the large UNIX family

▶ It was originally designed as a time-sharing system
▶ The main goal was to get the best throughput from the available hardware, by
making the best possible usage of resources (CPU, memory, I/O)
▶ Time determinism was not taken into account
▶ On the opposite, real-time constraints imply time determinism, even at the
expense of lower global throughput
▶ Best throughput and time determinism are contradictory requirements

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 438/520
 Linux and real-time approaches (1)

▶ Over time, two major approaches have been taken to bring real-time requirements
into Linux
▶ Approach 1
▶ Improve the Linux kernel itself so that it matches real-time requirements, by
providing bounded latencies, real-time APIs, etc.
▶ Approach taken by the mainline Linux kernel and the PREEMPT_RT project.
▶ Approach 2
▶ Add a layer below the Linux kernel that will handle all the real-time requirements, so
that the behavior of Linux doesn’t affect real-time tasks.
▶ Approach taken by RTLinux, RTAI and Xenomai

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 439/520
 Linux and real-time approaches (2)

An alternative approach is to use specific hardware to run real-time work on:

▶ Dedicating a CPU core to a real-time OS or to a real-time application, using some
kind of hypervizor.
▶ Running real-time work on an FPGA
▶ Running real-time work on a dedicated microcontroller. For example, the TI
AM335x CPU (used in the Beaglebone Black) has a ”Programmable Real-Time
Unit and Industrial Communication Subsystem (PRU-ICSS)”, which can be used
for real-time processing.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 440/520
 Real-time in embedded Linux systems

Improving the mainline Linux kernel with

PREEMPT_RT

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 441/520
 Understanding latency
▶ When developing real-time applications with a system such as Linux, the typical
scenario is the following
▶ An event from the physical world happens and gets notified to the CPU by means of
an interrupt
▶ The interrupt handler recognizes and handles the event, and then wake-up the user
space task that will react to this event
▶ Some time later, the user space task will run and be able to react to the physical
world event
▶ Real-time is about providing guaranteed worst case latencies for this reaction
time, called latency

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 442/520
 Linux kernel latency components

kernel latency = interrupt latency + handler duration + scheduler latency + scheduler

duration

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 443/520
 Interrupt latency

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 444/520
 Sources of interrupt latency
▶ One of the concurrency prevention mechanism used in the kernel is the spinlock
▶ It has several variants, but one of the variant commonly used to prevent
concurrent accesses between a process context and an interrupt context works by
disabling interrupts
▶ Critical sections protected by spinlocks, or other section in which interrupts are
explicitly disabled will delay the beginning of the execution of the interrupt
handler
▶ The duration of these critical sections is unbounded

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 445/520
 Interrupt handler duration

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 446/520
 Interrupt handler implementation

▶ In Linux, many interrupt handlers are split in two parts

▶ A top-half, started by the CPU as soon as interrupts are enabled. It runs with the
interrupt line disabled and is supposed to complete as quickly as possible.
▶ A bottom-half, scheduled by the top-half, which starts after all pending top-halves
have completed their execution.
▶ Therefore, for real-time critical interrupts, bottom-halves shouldn’t be used: their
execution is delayed by all other interrupts in the system.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 447/520
 Scheduler latency

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 448/520
 Understanding preemption (1)

▶ The Linux kernel is a preemptive operating system

▶ When a task runs in user space mode and gets interrupted by an interruption, if
the interrupt handler wakes up another task, this task can be scheduled as soon as
we return from the interrupt handler.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 449/520
 Understanding preemption (2)

▶ However, when the interrupt comes while the task is executing a system call, this
system call has to finish before another task can be scheduled.
▶ By default, the Linux kernel does not do kernel preemption.
▶ This means that the time before which the scheduler will be called to schedule
another task is unbounded.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 450/520
 Scheduler duration

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 451/520
 Sources of scheduler duration

▶ Before Linux 2.6, the execution time of the scheduler depended on the number of
processes on the system.
▶ Since Linux 2.6, the Linux kernel chooses the next process to run in constant time
(”O(1) scheduler” feature).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 452/520
 Other non-deterministic mechanisms

Outside of the critical path detailed previously, other non-deterministic mechanisms of

Linux can affect the execution time of real-time tasks
▶ Linux is highly based on virtual memory, as provided by an MMU, so that memory
is allocated on demand. Whenever an application accesses code or data for the
first time, it is loaded on demand, which can creates huge delays.
▶ Many C library services or kernel services are not designed with real-time
constraints in mind.
▶ To avoid such sources of non-determinism, your system should allocate all the
resources it needs ahead of time, before it is ready to react to events in a
real-time way. For the virtual memory needs, it will be done through the mlock()
and mlockall() system calls.
Issues can be addressed with a correct system design!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 453/520
 Priority inversion
A process with a low priority might hold a lock needed by a higher priority process,
effectively reducing the priority of this process. Things can be even worse if a middle
priority process uses the CPU.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 454/520
 Interrupt handler priority

In Linux, interrupt handlers are executed directly by the CPU interrupt mechanisms,
and not under control of the Linux scheduler. Therefore, all interrupt handlers have a
higher priority than all tasks running on the system.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 455/520
 The PREEMPT_RT project

▶ Long-term project lead by Linux kernel developers such as Thomas Gleixner,

Steven Rostedt and Ingo Molnar
▶ https://wiki.linuxfoundation.org/realtime/
▶ The goal is to gradually improve the Linux kernel regarding real-time requirements
and to get these improvements merged into the mainline kernel
▶ PREEMPT_RT development works very closely with the mainline development
▶ Many of the improvements designed, developed and debugged inside
PREEMPT_RT over the years are now part of the mainline Linux kernel
▶ The project is a long-term branch of the Linux kernel that ultimately should
disappear as everything will have been merged

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 456/520
 Improvements to the mainline kernel

Brought by PREEMPT_RT project since its beginning

▶ O(1) scheduler
▶ Kernel preemption
▶ Better POSIX real-time API support
▶ ftrace kernel function tracer
▶ Priority inheritance support for mutexes
▶ High-resolution timers
▶ Threaded interrupts
▶ Spinlock annotations

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 457/520
 Preemption options

Preemption models offered by the mainline kernel (Linux 5.7 status):

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 458/520
 1st option: no forced preemption

CONFIG_PREEMPT_NONE
Kernel code (interrupts, exceptions, system calls) never preempted. Default behavior in
standard kernels.
▶ Best for systems making intense computations, on which overall throughput is key.
▶ Best to reduce task switching to maximize CPU and cache usage (by reducing
context switching).
▶ Still benefits from some Linux real-time improvements: O(1) scheduler, increased
multiprocessor safety (work on RT preemption was useful to identify hard to find
SMP bugs).
▶ Can also benefit from a lower timer frequency (100 Hz instead of 250, 300 or
1000).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 459/520
 2nd option: voluntary kernel preemption

CONFIG_PREEMPT_VOLUNTARY
Kernel code can preempt itself
▶ Typically for desktop systems, for quicker application reaction to user input.
▶ Adds explicit rescheduling points (might_sleep()) throughout kernel code.
▶ Minor impact on throughput.
▶ Still used in: Ubuntu Desktop 20.04

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 460/520
 3rd option: preemptible kernel

CONFIG_PREEMPT
Most kernel code can be involuntarily preempted at any time. When a process
becomes runnable, no more need to wait for kernel code (typically a system call) to
return before running the scheduler.
▶ Exception: kernel critical sections (holding spinlocks). In a case you hold a
spinlock on a uni-processor system, kernel preemption could run another process,
which would loop forever if it tried to acquire the same spinlock.
▶ Typically for desktop or embedded systems with latency requirements in the
milliseconds range.
▶ Still a relatively minor impact on throughput.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 461/520
 Priority inheritance

▶ One classical solution to the priority inversion problem is called priority inheritance

▶ In the Linux kernel, mutexes support priority inheritance

▶ In user space, priority inheritance must be explicitly enabled on a per-mutex basis.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 462/520
 High resolution timers

▶ The resolution of the timers used to be bound to the resolution of the regular
system tick
▶ Usually 100 Hz or 250 Hz, depending on the architecture and the configuration
▶ A resolution of only 10 ms or 4 ms.
▶ Increasing the regular system tick frequency is not an option as it would consume
too many resources
▶ The high-resolution timers infrastructure allows to use the available hardware
timers to program interrupts at the right moment.
▶ Hardware timers are multiplexed, so that a single hardware timer is sufficient to
handle a large number of software-programmed timers.
▶ Usable directly from user space using the usual timer APIs

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 463/520
 Threaded interrupts

▶ To solve the interrupt inversion problem, PREEMPT_RT has introduced the

concept of threaded interrupts
▶ The interrupt handlers run in normal kernel threads, so that the priorities of the
different interrupt handlers can be configured
▶ The real interrupt handler, as executed by the CPU, is only in charge of masking
the interrupt and waking-up the corresponding thread
▶ The idea of threaded interrupts also allows to use sleeping spinlocks (see later)
▶ The conversion of interrupt handlers to threaded interrupts is not automatic:
drivers must be modified
▶ In PREEMPT_RT, all interrupt handlers are switched to threaded interrupts

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 464/520
 Real-time in embedded Linux systems

PREEMPT_RT patchset specifics

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 465/520
 New real-time preemption option (1)

The PREEMPT_RT patchset adds one new level of preemption, CONFIG_PREEMPT_RT:

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 466/520
 New real-time preemption option (2)

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 467/520
 CONFIG_PREEMPT_RT (1)

This level of preemption replaces all kernel spinlocks by mutexes (or so-called sleeping
spinlocks)
▶ Instead of providing mutual exclusion by disabling interrupts and preemption, they
are just normal locks: when contention happens, the process is blocked and
another one is selected by the scheduler.
▶ Works well with threaded interrupts, since threads can block, while usual interrupt
handlers could not.
▶ Some core, carefully controlled, kernel spinlocks remain as normal spinlocks
(spinlock annotations: differentiation now in mainline between spinning spinlocks
and those that can be converted to sleeping spinlocks).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 468/520
 CONFIG_PREEMPT_RT (2)

▶ With CONFIG_PREEMPT_RT, virtually all kernel code becomes preemptible

▶ An interrupt can occur at any time, when returning from the interrupt handler, the
woken up process can start immediately.
▶ The mechanism of threaded interrupts in PREEMPT_RT is still different from
the one merged in mainline
▶ In PREEMPT_RT, all interrupt handlers are unconditionally converted to threaded
interrupts.
▶ However, this is a temporary solution, until interesting drivers in mainline get
gradually converted to the threaded interrupt API.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 469/520
 Real-time in embedded Linux systems

Setting up PREEMPT_RT

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 470/520
 PREEMPT_RT setup (1)

▶ PREEMPT_RT is delivered as a patch against the mainline kernel

▶ Best to have a board supported by the mainline kernel, otherwise the
PREEMPT_RT patch may not apply and may require some adaptations
▶ Similarly, only long term stable releases of the kernel are supported, currently: 5.4,
4.19, 4.14, 4.9, 4.4. This is yet another advantage of using an LTS kernel release!
▶ Officially supported versions:
https://wiki.linuxfoundation.org/realtime/preempt_rt_versions
▶ Quick set up:
▶ Download the latest PREEMPT_RT patch
from https://kernel.org/pub/linux/kernel/projects/rt/
▶ Download and extract the corresponding mainline kernel version
▶ Apply the patch to the mainline kernel tree

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 471/520
 PREEMPT_RT setup (2)

▶ In the kernel configuration, be sure to enable

▶ CONFIG_PREEMPT_RT
▶ High-resolution timers
▶ Compile your kernel, and boot
▶ You are now running the real-time Linux kernel
▶ Of course, some system configuration remains to be done, in particular setting
appropriate priorities to the interrupt threads, which depend on your application.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 472/520
 PREEMPT_RT mainlining status (1)

▶ The project is making good progress since it got funding from the Linux
foundation in 2015.
▶ However, a reduction of the size of the PREEMPT_RT patchset is not really
visible yet:

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 473/520
 PREEMPT_RT mainlining status (2)

▶ However, the mainline Linux kernel is a moving target too, introducing new issues
for real-time (such as disabling preemption in BPF... see
https://lwn.net/Articles/802884/).
▶ A major achievement though in Linux 5.3: the CONFIG_PREEMPT_RT configuration
option is now in mainline, though it cannot be selected yet (missing
dependencies). This simplifies further mainlining work.
▶ See the latest status presentations:
▶ PREEMPT_RT: status and Q&A, Thomas Gleixner, Linux Plumbers 2019
Video: https://youtu.be/bpyFQJV5gCI?t=8489
▶ State of PREEMPT_RT - Sebastian Andrzej Siewior, Real-time Linux Summit 2019
https://bristot.me/wp-content/uploads/2019/11/rtls2019_07_rt_status.pdf

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 474/520
 Useful resources

About real-time support in the standard Linux kernel

▶ Real-Time is coming to Linux - What does that mean for you, Steven Rostedt,
VMWare, ELCE 2018
https://www.elinux.org/images/7/7e/Rostedt-elc-eu-2018-rt-what-does-
it-mean.pdf
Video: https://youtu.be/BxJm-Ujipcg
▶ Using SCHED_DEADLINE - Steven Rostedt, ELCE 2016
https://elinux.org/images/f/fe/Using_SCHED_DEADLINE.pdf
Video: https://youtu.be/TDR-rgWopgM
▶ The home page of the PREEMPT_RT project:
https://wiki.linuxfoundation.org/realtime/
▶ See also our books page.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 475/520
 Real-time in embedded Linux systems

Real-time application development with

PREEMPT_RT

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 476/520
 Development and compilation

▶ No special library is needed, the POSIX real-time API is part of the standard C
library
▶ The glibc C library is recommended, as support for some real-time features is not
mature in other C libraries
▶ Priority inheritance mutexes or NPTL on some architectures, for example
▶ Compile a program
▶ ARCH-linux-gcc -o myprog myprog.c -lrt
▶ To get the documentation of the POSIX API
▶ Install the manpages-posix-dev package
▶ Run man function-name

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 477/520
 Process, thread?

▶ Confusion about the terms process, thread and task

▶ In UNIX, a process is created using fork() and is composed of
▶ An address space, which contains the program code, data, stack, shared libraries, etc.
▶ One thread, that starts executing the main() function.
▶ Upon creation, a process contains one thread
▶ Additional threads can be created inside an existing process, using
pthread_create()
▶ They run in the same address space as the initial thread of the process
▶ They start executing a function passed as argument to pthread_create()

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 478/520
 Process, thread: kernel point of view

▶ The kernel represents each thread running in the system by a

struct task_struct structure.
▶ From a scheduling point of view, it makes no difference between the initial thread
of a process and all additional threads created dynamically using
pthread_create()

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 479/520
 Creating threads

▶ Linux support the POSIX thread API

▶ To create a new thread
pthread_create(pthread_t *thread, pthread_attr_t *attr,
void *(*routine)(void*), void *arg);
▶ The new thread will run in the same address space, but will be scheduled
independently
▶ Exiting from a thread

pthread_exit(void *value_ptr);
▶ Waiting for the termination of a thread

pthread_join(pthread_t *thread, void **value_ptr);

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 480/520
 Scheduling classes (1)

The Linux kernel scheduler support different scheduling classes

▶ The default class (SCHED_OTHER), in which processes are started by default, is a
time-sharing class
▶ All processes, regardless of their priority, get some CPU time
▶ The proportion of CPU time they get is dynamic and affected by the nice value,
which ranges from -20 (highest) to 19 (lowest). Can be set using the nice or
renice commands

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 481/520
 Scheduling classes (2)

▶ The real-time classes SCHED_FIFO, SCHED_RR and SCHED_DEADLINE

▶ The highest RT priority process gets all the CPU time, until it blocks.
▶ With SCHED_FIFO, First In, First Out, each additional process with the same RT
priority has to wait for the first ones to block.
▶ In SCHED_RR, Round-Robin scheduling between the processes with the same RT
priority. All must block before lower priority processes get CPU time.
▶ In SCHED_DEADLINE: guarantees that an RT task will be given an exact amount of
cpu time every period.
▶ RT Priorities ranging from 0 (lowest) to 99 (highest)

See https://man7.org/linux/man-pages/man7/sched.7.html for details.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 482/520
 Using scheduling classes (1)

▶ An existing program can be started in a specific scheduling class with a specific

priority using the chrt command line tool
▶ Example: chrt -f 99 ./myprog
-f: SCHED_FIFO
-r: SCHED_RR
-d: SCHED_DEADLINE
▶ The sched_setscheduler() API can be used to change the scheduling class and
priority of a process

int sched_setscheduler(pid_t pid, int policy,

const struct sched_param *param);

▶ policy can be SCHED_OTHER, SCHED_FIFO, SCHED_RR, SCHED_DEADLINE, etc. (others

exist).
▶ param is a structure containing the priority

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 483/520
 Using scheduling classes (2)

▶ The priority can be set on a per-thread basis when a thread is created

struct sched_param parm;

pthread_attr_t attr;

pthread_attr_init(&attr);
pthread_attr_setinheritsched(&attr, PTHREAD_EXPLICIT_SCHED);
pthread_attr_setschedpolicy(&attr, SCHED_FIFO);
parm.sched_priority = 42;
pthread_attr_setschedparam(&attr, &parm);

▶ Then the thread can be created using pthread_create(), passing the attr
structure.
▶ Several other attributes can be defined this way: stack size, etc.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 484/520
 Memory locking

▶ In order to solve the non-determinism introduced by virtual memory, memory can

be locked
▶ Guarantee that the system will keep it allocated
▶ Guarantee that the system has pre-loaded everything into memory
▶ mlockall(MCL_CURRENT | MCL_FUTURE);
▶ Locks all the memory of the current address space, for currently mapped pages and
pages mapped in the future
▶ Other, less useful parts of the API: munlockall, mlock, munlock.
▶ Watch out for non-currently mapped pages
▶ Stack pages
▶ Dynamically-allocated memory

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 485/520
 Mutexes

▶ Allows mutual exclusion between two threads in the same address space
▶ Initialization/destruction

pthread_mutex_init(pthread_mutex_t *mutex,
const pthread_mutexattr_t *mutexattr);
pthread_mutex_destroy(pthread_mutex_t *mutex);

▶ Lock/unlock

pthread_mutex_lock(pthread_mutex_t *mutex);
pthread_mutex_unlock(pthread_mutex_t *mutex);

▶ Priority inheritance must be activated explicitly

pthread_mutexattr_t attr;
pthread_mutexattr_init (&attr);
pthread_mutexattr_setprotocol(&attr, PTHREAD_PRIO_INHERIT);

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 486/520
 Timers

▶ Timer creation

timer_create(clockid_t clockid, struct sigevent *evp,

timer_t *timerid);

▶ clockid is usually CLOCK_MONOTONIC. sigevent defines what happens upon timer

expiration: send a signal or start a function in a new thread. timerid is the returned
timer identifier.
▶ Configure the timer for expiration at a given time

timer_settime(timer_t timerid, int flags,

struct itimerspec *newvalue,
struct itimerspec *oldvalue);

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 487/520
 Timers (2)

▶ Delete a timer
timer_delete(timer_t timerid)

▶ Get the resolution of a clock, clock_getres

▶ Other functions: timer_getoverrun(), timer_gettime()

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 488/520
 Signals

▶ Signals are asynchronous notification mechanisms

▶ Notification occurs either
▶ By the call of a signal handler. Be careful with the limitations of signal handlers!
▶ By being unblocked from the sigwait(), sigtimedwait() or sigwaitinfo()
functions. Usually better.
▶ Signal behavior can be configured using sigaction()
▶ The mask of blocked signals can be changed with pthread_sigmask()
▶ Delivery of a signal using pthread_kill() or tgkill()
▶ All signals between SIGRTMIN and SIGRTMAX, 32 signals under Linux.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 489/520
 Inter-process communication

▶ Semaphores
▶ Usable between different processes using named semaphores
▶ sem_open(), sem_close(), sem_unlink(), sem_init(), sem_destroy(),
sem_wait(), sem_post(), etc.
▶ Message queues
▶ Allows processes to exchange data in the form of messages.
▶ mq_open(), mq_close(), mq_unlink(), mq_send(), mq_receive(), etc.
▶ Shared memory
▶ Allows processes to communicate by sharing a segment of memory
▶ shm_open(), ftruncate(), mmap(), munmap(), close(), shm_unlink()

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 490/520
 Real-time in embedded Linux systems

Debugging latencies in PREEMPT_RT

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 491/520
 ftrace - Kernel function tracer

Infrastructure that can be used for debugging or analyzing latencies and performance
issues in the kernel.
▶ Very well documented in Documentation/trace/ftrace.txt
▶ Negligible overhead when tracing is not enabled at run-time.
▶ Can be used to trace any kernel function!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 492/520
 Using ftrace

▶ Tracing information available through the tracefs virtual fs

▶ Mount this filesystem as follows:
mount -t tracefs nodev /sys/kernel/tracing
▶ Check available tracers in /sys/kernel/tracing/available_tracers

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 493/520
 Scheduling latency tracer
CONFIG_SCHED_TRACER (Kernel Hacking section)
▶ Maximum recorded time between waking up a top priority task and its scheduling on a CPU,
expressed in us.
▶ Check that wakeup is listed in /sys/kernel/tracing/available_tracers
▶ To select, reset and enable this tracer:

echo wakeup > /sys/kernel/tracing/current_tracer

echo 0 > /sys/kernel/tracing/tracing_max_latency
echo 1 > /sys/kernel/tracing/tracing_enabled
▶ Let your system run, in particular real-time tasks.
Dummy example: chrt -f 5 sleep 1
▶ Disable tracing:

echo 0 > /sys/kernel/tracing/tracing_enabled

▶ Read the maximum recorded latency and the corresponding trace:

cat /sys/kernel/tracing/tracing_max_latency

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 494/520
 Real-time in embedded Linux systems

Real-time extensions to the Linux kernel

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 495/520
 Linux real-time extensions

Three generations A common principle

▶ RTLinux ▶ Add an extra layer between the
▶ RTAI hardware and the Linux kernel, to
▶ Xenomai manage real-time tasks separately.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 496/520
 RTLinux

First real-time extension for Linux, created by Victor Yodaiken.

▶ Nice, but the author filed a software patent covering the addition of real-time
support to general operating systems as implemented in RTLinux!
▶ Its Open Patent License drew many developers away and frightened users.
Community projects like RTAI and Xenomai now attract most developers and
users.
▶ February, 2007: RTLinux rights sold to Wind River. Today, no longer advertised
by Wind River.
▶ Project completely dead.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 497/520
 RTAI

https://www.rtai.org/ - Real-Time Application Interface for Linux

▶ Created in 1999, by Prof. Paolo Mantegazza (long time contributor to RTLinux),
Dipartimento di Ingegneria Aerospaziale Politecnico di Milano (DIAPM).
▶ Community project. Significant user base. Attracted contributors frustrated by
the RTLinux legal issues.
▶ Only really actively maintained on x86
▶ May offer slightly better latencies than Xenomai, at the expense of a less
maintainable and less portable code base
▶ Since RTAI is not really maintained on ARM and other embedded architectures,
our presentation is focused on Xenomai.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 498/520
 Xenomai project

https://www.xenomai.org/
▶ Started in 2001 as a project aiming at emulating traditional
RTOS.
▶ Initial goals: facilitate the porting of programs to GNU / Linux.
▶ Skins mimicking the APIs of the VxWorks and pSOS traditional
real-time operating systems, as well as the POSIX API, and
offering a “native” API too.
▶ Offers two solutions:
▶ Cobalt: (like Co-kernel), using a micro-kernel dedicated to
running real-time tasks. This will never be merged in the mainline
kernel.
▶ Mercury: (like Merge), based on the mainline kernel with
PREEMPT_RT.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 499/520
 Xenomai Cobalt architecture

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 500/520
 Xenomai user space real-time support

▶ Supported architectures are the ones supported by the ipipe patch: arm, arm64,
blackfin, powerpc, x86
▶ Two modes are defined for a thread
▶ the primary mode, where the thread is handled by the Xenomai scheduler
▶ the secondary mode, when it is handled by the Linux scheduler.
▶ Thanks to the services of the Adeos I-pipe service, Xenomai system calls are
defined.
▶ A thread migrates from secondary mode to primary mode when such a system call is
issued
▶ It migrates from primary mode to secondary mode when a Linux system call is
issued, or to handle gracefully exceptional events such as exceptions or Linux signals.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 501/520
 Life of a Xenomai application

▶ Xenomai applications are started like normal Linux processes, they are initially
handled by the Linux scheduler and have access to all Linux services
▶ After their initialization, they declare themselves as real-time applications, which
migrates them to primary mode. In this mode:
▶ They are scheduled directly by the Xenomai scheduler, so they have the real-time
properties offered by Xenomai
▶ They don’t have access to any Linux service, otherwise they get migrated back to
secondary mode and loose all real-time properties
▶ They can only use device drivers that are implemented in Xenomai, not the ones of
the Linux kernel
▶ Need to implement device drivers in Xenomai, and to split real-time and non
real-time parts of your applications.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 502/520
 Real Time Driver Model (RTDM)

▶ An approach to unify the interfaces for developing device drivers and associated
applications under real-time Linux
▶ An API very similar to the native Linux kernel driver API
▶ Allows to develop in kernel space:
▶ Character-style device drivers
▶ Network-style device drivers
▶ Current notable RTDM based drivers:
▶ Serial port controllers;
▶ RTnet UDP/IP stack;
▶ RT socket CAN, drivers for CAN controllers;
▶ Analogy, fork of the Comedi project, drivers for acquisition cards.
https://xenomai.org/documentation/xenomai-
3/html/xeno3prm/group__rtdm.html

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 503/520
 Real-time in embedded Linux systems

Setting up Xenomai

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 504/520
 How to build the Xenomai kernel (Cobalt)

▶ Download Xenomai sources at

https://xenomai.org/downloads/xenomai/stable/latest/
▶ Download the latest i-pipe patch for your architecture:
https://xenomai.org/downloads/ipipe/
▶ Download the corresponding kernel.org version
▶ Xenomai offers a script to patch your kernel. Example (arm):
./scripts/prepare-kernel.sh \
--linux=../linux-4.19.124 \
--ipipe=../ipipe-core-4.19.124-cip27-arm-09.patch
--arch=arm \
▶ Run the kernel configuration menu.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 505/520
 Linux options for Xenomai configuration

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 506/520
 How to build Xenomai user space

▶ User space libraries are compiled using the traditional autotools

▶ ./configure --host=arm-
linux && make && make DESTDIR=/your/rootfs/ install
▶ Xenomai installs pkg-config files which helps you to compile your own programs
against the Xenomai libraries.
▶ See Xenomai’s examples directory.
▶ Installation details may be found in the README.INSTALL guide.
▶ Build systems such as Buildroot can compile the Xenomai user space for you.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 507/520
 Real-time in embedded Linux systems

Developing applications on Xenomai

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 508/520
 The POSIX skin

▶ The POSIX skin allows to recompile without changes a traditional POSIX

application so that instead of using Linux real-time services, it uses Xenomai
services
▶ https:
//www.xenomai.org/index.php/Porting_POSIX_applications_to_Xenomai
▶ Clocks and timers, condition variables, message queues, mutexes, semaphores,
shared memory, signals, thread management
▶ Good for existing code or programmers familiar with the POSIX API
▶ Of course, if the application uses any Linux service that isn’t available in
Xenomai, it will switch back to secondary mode
▶ To link an application against the POSIX skin

CFL=`pkg-config --cflags libxenomai_posix`

LDF=`pkg-config --libs libxenomai_posix`
ARCH-gcc $CFL -o rttest rttest.c $LDF
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 509/520
 Communication with a normal task

▶ If a Xenomai real-time application using the POSIX skin wishes to communicate

with a separate non-real-time application, it must use the rtipc mechanism
▶ In the Xenomai application, create an IPCPROTO_XDDP socket

socket(AF_RTIPC, SOCK_DGRAM, IPCPROTO_XDDP);

setsockopt(s, SOL_RTIPC, XDDP_SETLOCALPOOL,
&poolsz, sizeof(poolsz));
memset(&saddr, 0, sizeof(saddr));
saddr.sipc_family = AF_RTIPC;
saddr.sipc_port = PORTX;
ret = bind(s, (struct sockaddr *)&saddr, sizeof(saddr));

▶ And then the normal socket API sendto() / recvfrom()

▶ In the Linux application
▶ Open /dev/rtpPORTX, where PORTX is the XDDP port
▶ Use read() and write()

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 510/520
 Alchemy: the native API

Xenomai proposes its own API for developing real-time tasks

▶ https://xenomai.org/documentation/xenomai-
3/html/xeno3prm/group__alchemy.html
▶ Xenomai developers find it a more coherent and more flexible API than the
POSIX API, and easier to learn and understand.
▶ However, beware that programming with the POSIX API is much more portable
(ability to use regular Linux, Linux with PREEMPT_RT, Xenomai and other
POSIX operating systems).

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 511/520
 Benchmarks: mainline vs PREEMPT_RT vs Xenomai
cyclictest results on BeagleBone Black (ARM), with different kernel setups.
Wake-up time measured in µs.

Table: Idle CPU

min avg max
Mainline Linux 39 43 1046
PREEMPT_RT 16 21 52
Xenomai 2 8 9 41
Xenomai 3 7 7 35

Table: Stressed CPU

min avg max
Mainline Linux 39 52 1097
PREEMPT_RT 18 25 62
Xenomai 2 8 19 45
Xenomai 3 6 16 37

http://wiki.csie.ncku.edu.tw/embedded/xenomai/rtlws_paper.pdf
- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 512/520
 Organizations

▶ https://www.osadl.org
Open Source Automation Development Lab (OSADL)
Targets machine and plant control systems. Most member companies are German
(Thomas Gleixner is on board). Supports the use of PREEMPT_RT and Xenomai
and contributes to these projects. Shares useful documentation and resources.
They also organize a yearly Real Time Linux Workshop.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 513/520
 Practical lab - Real-time - Scheduling latency

▶ Check clock accuracy.

▶ Start processes with real-time priority.
▶ Build a real-time application against the
standard POSIX real-time API, and against
Xenomai’s POSIX skin
▶ Compare scheduling latency on your system,
between a standard kernel, a kernel with
PREEMPT_RT and a kernel with Xenomai.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 514/520
 References

References

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 515/520
 Books

▶ Mastering Embedded Linux, Second Edition, Packt Publishing

By Chris Simmonds, June 2017
An up-to-date resource covering most aspects of embedded Linux
development.
https://bit.ly/2A9Pb5Y
▶ Embedded Linux Primer, Second Edition, Prentice Hall
By Christopher Hallinan, October 2010
Covers a very wide range of interesting topics.
https://j.mp/17NYxBP
▶ Embedded Linux System Design and Development
P. Raghavan, A. Lad, S. Neelakandan, Auerbach, Dec. 2005. Very good
coverage of the POSIX programming API (still up to date).
https://j.mp/19X8iu2

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 516/520
 Web sites

▶ ELinux.org, https://elinux.org, a Wiki entirely dedicated to embedded Linux.

Lots of topics covered: real-time, filesystem, multimedia, tools, hardware
platforms, etc. Interesting to explore to discover new things.
▶ LWN, https://lwn.net, very interesting news site about Linux in general, and
specifically about the kernel. Weekly edition, available for free after one week for
non-paying visitors.
▶ Linux Gizmos, https://linuxgizmos.com, a news site about embedded Linux,
mainly oriented on hardware platforms related news.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 517/520
 International conferences

Useful conferences featuring embedded Linux and kernel topics

▶ Embedded Linux Conference:

https://embeddedlinuxconference.com/
Organized by the Linux Foundation: USA (February-April), in Europe
(October-November). Very interesting kernel and user space topics for embedded
systems developers. Presentation slides and videos freely available
▶ Linux Plumbers, https://linuxplumbersconf.org
Conference on the low-level plumbing of Linux: kernel, audio, power management,
device management, multimedia, etc.
▶ FOSDEM: https://fosdem.org (Brussels, February)
For developers. Presentations about system development.

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 518/520
 Last slides

Last slides

© Copyright 2004-2020, Bootlin.

Creative Commons BY-SA 3.0 license.
embedded Linux and kernel engineering
Corrections, suggestions, contributions and translations are welcome!

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 519/520
 Last slide

Thank you!
And may the Source be with you

- Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 520/520