Downloading - All IT eBooks

www.allitebooks.com
Embedded Linux Development
with Yocto Project
Develop fascinating Linux-based projects using the
groundbreaking Yocto Project tools
Otavio Salvador
Daiane Angolini
BIRMINGHAM - MUMBAI
www.allitebooks.com
Embedded Linux Development with Yocto Project
Copyright © 2014 Packt Publishing
All rights reserved. No part of this book may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, without the prior written
permission of the publisher, except in the case of brief quotations embedded in
critical articles or reviews.
Every effort has been made in the preparation of this book to ensure the accuracy
of the information presented. However, the information contained in this book is
sold without warranty, either express or implied. Neither the authors, nor Packt
Publishing, and its dealers and distributors will be held liable for any damages
caused or alleged to be caused directly or indirectly by this book.
Packt Publishing has endeavored to provide trademark information about all of the
companies and products mentioned in this book by the appropriate use of capitals.
However, Packt Publishing cannot guarantee the accuracy of this information.
First published: July 2014
Production reference: 1020714
Published by Packt Publishing Ltd.
Livery Place
35 Livery Street
Birmingham B3 2PB, UK.
ISBN 978-1-78328-233-3
www.packtpub.com
Cover image by Jarek Blaminsky (milak6@wp.pl)
www.allitebooks.com
Credits
Authors
Project Coordinators
Otavio Salvador
Harshal Ved
Daiane Angolini
Sageer Parkar
Reviewers
Proofreaders
Radek Dostál
Simran Bhogal
Alex González
Stephen Copestake
Rogerio Nunes
Amy Guest
Jeffrey Osier-Mixon
Alexandru Vaduva
Commissioning Editor
Aarthi Kumaraswamy
Acquisition Editor
Harsha Bharwani
Indexers
Hemangini Bari
Mariammal Chettiyar
Tejal Soni
Graphics
Sheetal Aute
Valentina Dsilva
Content Development Editor
Sankalp Pawar
Abhinash Sahu
Production Coordinator
Technical Editor
Alwin Roy
Manan Badani
Cover Work
Copy Editor
Sayanee Mukherjee
Alwin Roy
www.allitebooks.com
About the Authors
Otavio Salvador loves to play video games and started his free software activities
in 1999. In 2002, he founded O.S. Systems, a company focused on embedded system
development services and consultancy worldwide, creating and maintaining
customized BSPs and helping companies with their release management challenges.
This resulted in him joining the OpenEmbedded community in 2008, when he became
an active contributor to the OpenEmbedded project, culminating in his attribution as
the maintainer of the Freescale ARM BSP layer in the Yocto Project in 2011.
Daiane Angolini has been focusing on embedded technologies for the past 8 years.
Since 2008, she has been working on Freescale Semiconductors as an application
engineer, on internal development and porting custom applications from Android
to Freescale architectures, and on customer support for ARM processors of the i.MX
family, while also participating in Freescale forums. She has been working with the
Yocto Project tools through meta-fsl-arm, the BSP meta layer that provides board
support for Freescale ARM machines, since 2012. The desire to become an expert in ice
cream making has been keeping her busy in her spare time for the past year.
We initially want to thank our families. They provided lovely
support and helped us to get on track for this project.
This project has only been possible because we had support from
many people who provided insights, reviews, material, and
guidance during the full period of conception and production of
this book. We'd like to give special thanks to (in alphabetic order):
Alex González, Alexandru Vaduva, Harsha Bharwani,
Jeffrey Osier-Mixon, John Weber, Manan Badani, Paul Eggleton,
Rogerio Nunes, Radek Dostál, Sageer Parkar, and Sankalp Pawar
- Otavio Salvador and Daiane Angolini
www.allitebooks.com
About the Reviewers
Radek Dostál is a fan of Linux and has been using it for the last 15 years. During
his exchange studies in the US, he acquired a passion for embedded systems, and
combining Linux with embedded systems has been his bread and butter ever since.
The Yocto Project has had a great impact on Radek's work; he managed to persuade his
team and managers to switch to Yocto Project for an important project, thus building a
solid foundation for several successful follow-up projects. Radek likes to contribute to
open source projects as part of his work as well as during his free time. However, if the
weather is good during the weekend, you are most likely to find him in the mountains.
Alex González is Principal Software Engineer at Digi International and one of the
maintainers of Digi Embedded Yocto distribution.
He started working professionally with embedded systems in 1999 and the Linux
kernel in 2004, designing products for next-generation IP networks in the UK
start-up scene, and he followed his interests into M2M and the Internet of Things.
Born and raised in Spain, he has his second home in the UK, where he lived for
over 10 years and received his MSc degree in Communication Systems from the
University of Portsmouth. Alex currently lives in La Rioja, Spain, where he is known
to enjoy photography and a good Riojan wine.
www.allitebooks.com
Rogerio Nunes has over 11 years of experience in embedded systems. He received
his MS (2009) and BS (2004) in Electrical Engineering from the University of São
Paulo in Brazil, where he also worked for 8 years in research and development. In
2011, Rogerio started his career at Freescale in São Paulo as a Field Applications
Engineer (FAE), supporting high-end multimedia SoCs. Later in 2012, Rogerio
moved to Boston in the same role as FAE for the same company. Rogerio's fields of
expertise include digital TV, multimedia, video coding, and software development.
In these areas, Rogerio has developed systems with different technologies (from
VHDL to high-level software), and he has also led development teams.
Jeffrey Osier-Mixon is a Yocto Project community manager. He has been working
directly with Linux since the late 1990s and with embedded systems and open source
software for over 20 years, most often as a technical writer and editor, freelance
writer and journalist, and community manager. He has been a regular speaker at
open source conferences worldwide since 2008. He is employed by Intel Corporation
to help the Yocto Project succeed.
I would like to thank Otavio and Daiane for writing this book and
giving me the opportunity to review it. I am sure it will be a valuable
asset to the Yocto Project community.
Alexandru Vaduva is an embedded Linux software engineer, focused on open
source and free software. He has an inquisitive mind and also believes that actions can
speak more about someone than their own words. He is a strong supporter of the idea
that there is no need to reinvent the wheel, but there is always room for improvement.
His knowledge background includes C, Yocto, Linux, Bash, and Python, but he is
also open to trying new things and testing new technologies.
Big thanks to all the people who believed in me, and the open source
communities that helped me evolve and kept me motivated.
www.allitebooks.com
www.PacktPub.com
Support files, eBooks, discount offers, and more
You might want to visit www.PacktPub.com for support files and downloads related
to your book.
Did you know that Packt offers eBook versions of every book published, with PDF
and ePub files available? You can upgrade to the eBook version at www.PacktPub.
com and as a print book customer, you are entitled to a discount on the eBook copy.
Get in touch with us at service@packtpub.com for more details.
At www.PacktPub.com, you can also read a collection of free technical articles, sign
up for a range of free newsletters and receive exclusive discounts and offers on Packt
books and eBooks.
TM
http://PacktLib.PacktPub.com
Do you need instant solutions to your IT questions? PacktLib is Packt's online
digital book library. Here, you can access, read and search across Packt's entire
library of books.
Why subscribe?
• Fully searchable across every book published by Packt
• Copy and paste, print and bookmark content
• On demand and accessible via web browser
Free access for Packt account holders
If you have an account with Packt at www.PacktPub.com, you can use this to access
PacktLib today and view nine entirely free books. Simply use your login credentials
for immediate access.
www.allitebooks.com
www.allitebooks.com
Table of Contents
Preface1
Chapter 1: Meeting the Yocto Project
7
What is the Yocto Project?
7
Delineating the Yocto Project
8
Understanding Poky
8
Using BitBake
9
OpenEmbedded-Core9
Metadata10
The alliance of OpenEmbedded Project and Yocto Project
10
Summary11
Chapter 2: Baking Our Poky-based System
13
Chapter 3: Using Hob to Bake an Image
21
Chapter 4: Grasping the BitBake Tool
27
Configuring a host system
13
Installing Poky on Debian
14
Installing Poky on Fedora
14
Downloading the Poky source code
14
Preparing the build environment
15
Knowing the local.conf file
16
Building a target image
17
Running images in QEMU
19
Summary20
Building an image using Hob
Customizing an image with Hob
Summary
Understanding the BitBake tool
Exploring metadata
www.allitebooks.com
21
23
26
27
28
Table of Contents
Parsing metadata
28
Dependencies29
Preferring and providing recipes
30
Fetching the source code
31
Remote file downloads
31
Git repositories
32
Other repositories
33
Optimizing the source code download
33
Disabling network access
35
Understanding BitBake's tasks
35
Extending tasks
37
Generating a root filesystem image
37
Summary
39
Chapter 5: Detailing the Temporary Build Directory
41
Chapter 6: Assimilating Packaging Support
49
Chapter 7: Diving into BitBake Metadata
59
Detailing the build directory
41
Constructing the build directory
42
Exploring the temporary build directory
42
Understanding the work directory
43
Understanding the sysroot directories
46
Summary47
Using supported package formats
49
List of supported package formats
49
Choosing a package format
50
Running code during package installation
51
Understanding shared state cache
53
Explaining package versioning
54
Package feeds
55
Using package feeds
56
Summary57
Using metadata
Working with metadata
The basic variable setting
Variable expansion
Setting a default value using ?=
Setting a default value using ??=
Immediate variable expansion
Appending and prepending
Conditional metadata set
[ ii ]
59
60
60
60
61
61
61
62
63
Table of Contents
Conditional appending
File inclusion
Python variable expansion
Defining executable metadata
Defining Python functions in the global namespace
The inheritance system
Summary
63
63
64
64
64
65
65
Chapter 8: Developing with the Yocto Project
67
Chapter 9: Debugging with the Yocto Project
73
Chapter 10: Exploring External Layers
81
Chapter 11: Creating Custom Layers
87
Deciphering the software development kit
67
Working with the Poky SDK
68
Using an image-based SDK
68
Generic SDK – meta-toolchain
69
Using a SDK
70
Developing applications on the target
70
Integrating with Eclipse
72
Summary72
Differentiating metadata and application debugging
73
Tracking image, package, and SDK contents
74
Debugging packaging
75
Logging information during task execution
76
Utilizing a development shell
77
Using the GNU Project Debugger for debugging
78
Summary79
Powering flexibility with layers
Detailing the layer's source code
Adding meta layers
Summary
Making a new layer
Adding metadata to the layer
Creating an image
Adding a package recipe
Writing a machine definition
Using a custom distribution
Machine features versus distro features
Understanding the variables scope
Summary
[ iii ]
81
83
85
86
87
89
89
91
92
93
95
96
96
Table of Contents
Chapter 12: Customizing Existing Recipes
97
Common use cases
97
Adding extra options to recipes based on Autoconf
98
Applying a patch
98
Adding extra files to the existing packages
99
Understanding file searching paths
99
Changing recipe feature configuration
100
Customizing BusyBox
101
Customizing the linux-yocto framework
101
Summary102
Chapter 13: Achieving GPL Compliance
103
Chapter 14: Booting Our Custom Embedded Linux
111
Understanding copyleft
103
Copyleft compliance versus proprietary code
104
Some guidelines for license compliance
104
Managing software licensing with Poky
104
Commercial licenses
105
Using Poky to achieve copyleft compliance
106
License auditing
106
Providing the source code
107
Providing compilation scripts and source code modifications
108
Providing license text
108
Summary109
Exploring the Wandboard
Discovering Freescale ARM BSP
Using Wandboard with the Yocto Project
Building an image for Wandboard
Booting Wandboard from the SD card
Summary
112
113
113
115
115
116
Appendix: References
117
Index119
[ iv ]
Preface
Considering the current technology trend, Linux is the next big thing. Linux has
consistently released cutting-edge open source products, and embedded systems
have been added to the technological portfolio of mankind.
The Yocto Project is in an optimal position to be the choice for your projects; it
provides a rich set of tools to help you to use most of your energy and resources in
your product development, instead of reinventing the wheel.
The usual tasks and requirements for embedded Linux-based products and
development teams were the guidelines for this book's conception. Written by active
community members with a practical and straightforward approach, it is a stepping
stone for both your learning curve and your product's project.
What this book covers
Chapter 1, Meeting the Yocto Project, presents the history of the Yocto Project,
showing the parts that compose it.
Chapter 2, Baking Our Poky-based System, introduces the environment needed for
the first build.
Chapter 3, Using Hob to Bake an Image, shows the user-friendly graphical interface
that can be used as a wrapper for configuration and as a build tool.
Chapter 4, Grasping the BitBake Tool, presents the first concepts and premises of the
tool used to control all other pieces of the Yocto Project.
Chapter 5, Detailing the Temporary Build Directory, details the output directory tree of a
build with focus on the tmp directory.
Chapter 6, Assimilating Packaging Support, introduces the package concepts and details
the packaging support used by the Yocto Project.
Preface
Chapter 7, Diving into BitBake Metadata, details the concepts and syntaxes used by the
Yocto Project metadata, both in recipes and configuration files.
Chapter 8, Developing with the Yocto Project, details how to use the Yocto Project to
generate a custom development environment.
Chapter 9, Debugging with the Yocto Project, details which debug tools the Yocto Project
provides and how to use them.
Chapter 10, Exploring External Layers, explores one of the most important concepts of
the Yocto Project, which is the flexibility of using external layers.
Chapter 11, Creating Custom Layers, practices the steps of creation of layers.
Chapter 12, Customizing Existing Recipes, lists the common use cases of recipe
customization and how to achieve them properly.
Chapter 13, Achieving GPL Compliance, summarizes the tasks and concepts involved in
a copyleft compliance product.
Chapter 14, Booting Our Custom Embedded Linux, uses a real hardware machine
together with the Yocto Project's tools.
Appendix, References, lists the references used in the book.
What you need for this book
To better understand this book, it is important that you have some previous
background in some topics that are not covered or are just briefly mentioned along
the text.
A basic understanding of the GNU/Linux environment usage and embedded Linux
is important along with general concepts used in development as compilation,
debugging, deployment, and installation. Some experience with Shell Script and
Python is a bonus because these programming languages are core technologies used
extensively by the Yocto Project's tools.
However, the concepts enumerated should not discourage you from reading this
book as they can be learned concurrently.
[2]
Preface
Who this book is for
This book is intended to be read by engineers and enthusiasts with embedded Linux
experience willing to learn the Yocto Project's tools for evaluation, comparison, or
use in a project. This book is aimed to get you up to speed quickly and to prevent
you from getting trapped by the usual learning curve pitfalls.
Conventions
In this book, you will find a number of styles of text that distinguish between
different kinds of information. Here are some examples of these styles, and an
explanation of their meaning.
Code words in text, database table names, folder names, filenames, file extensions,
pathnames, dummy URLs, user input, and Twitter handles are shown as follows:
"Throughout the book, we will use build as the build directory."
A block of code is set as follows:
BB_NUMBER_THREADS ?= "${@oe.utils.cpu_count()}"
PARALLEL_MAKE ?= "-j ${@oe.utils.cpu_count()}"
MACHINE ??= "qemux86"
Any command-line input or output is written as follows:
$: sudo apt-get install gawk wget git-core diffstat unzip texinfo
build-essential chrpath
New terms and important words are shown in bold. Words that you see on the
screen, in menus or dialog boxes for example, appear in the text like this: "If we plan
to build a standard image, we can click on Build Image and wait for BitBake to run
the required tasks to build it."
Warnings or important notes appear in a box like this.
Tips and tricks appear like this.
[3]
Preface
Reader feedback
Feedback from our readers is always welcome. Let us know what you think about
this book—what you liked or may have disliked. Reader feedback is important for us
to develop titles that you really get the most out of.
To send us general feedback, simply send an e-mail to feedback@packtpub.com,
and mention the book title via the subject of your message.
If there is a topic that you have expertise in and you are interested in either writing
or contributing to a book, see our author guide on www.packtpub.com/authors.
Customer support
Now that you are the proud owner of a Packt book, we have a number of things to
help you to get the most from your purchase.
Downloading the color images of the book
We also provide you a PDF file that has color images of the screenshots/diagrams
used in this book. The color images will help you better understand the changes in
the output. You can download this file from: https://www.packtpub.com/sites/
default/files/downloads/2333OS_ColoredImages.pdf
Errata
Although we have taken every care to ensure the accuracy of our content, mistakes
do happen. If you find a mistake in one of our books—maybe a mistake in the text or
the code—we would be grateful if you would report this to us. By doing so, you can
save other readers from frustration and help us improve subsequent versions of this
book. If you find any errata, please report them by visiting http://www.packtpub.
com/submit-errata, selecting your book, clicking on the errata submission form link,
and entering the details of your errata. Once your errata are verified, your submission
will be accepted and the errata will be uploaded on our website, or added to any list of
existing errata, under the Errata section of that title. Any existing errata can be viewed
by selecting your title from http://www.packtpub.com/support.
[4]
Preface
Piracy
Piracy of copyright material on the Internet is an ongoing problem across all media.
At Packt, we take the protection of our copyright and licenses very seriously. If you
come across any illegal copies of our works, in any form, on the Internet, please
provide us with the location address or website name immediately so that we can
pursue a remedy.
Please contact us at copyright@packtpub.com with a link to the suspected
pirated material.
We appreciate your help in protecting our authors, and our ability to bring you
valuable content.
Questions
You can contact us at questions@packtpub.com if you are having a problem with
any aspect of the book, and we will do our best to address it.
[5]
Meeting the Yocto Project
In this chapter, we will be introduced to the Yocto Project. The main concepts of the
project, which are constantly used throughout the book, are discussed here. We will
discuss the Yocto Project history, OpenEmbedded, Poky, BitBake, and Metadata in
brief, so fasten your seat belt and welcome aboard!
What is the Yocto Project?
The Yocto Project is a Linux Foundation workgroup defined as:
"The Yocto Project provides open source, high-quality infrastructure and tools
to help developers create their own custom Linux distributions for any hardware
architecture, across multiple market segments. The Yocto Project is intended to
provide a helpful starting point for developers."
The Yocto Project is an open source collaboration project that provides templates,
tools, and methods to help us create custom Linux-based systems for embedded
products regardless of the hardware architecture. Being managed by a Linux
Foundation fellow, the project remains independent of its member organizations that
participate in various ways and provide resources to the project.
It was founded in 2010 as a collaboration of many hardware manufacturers, open
source operating systems, vendors, and electronics companies in an effort to reduce
their work duplication, providing resources and information catering to both new
and experienced users.
Among these resources is OpenEmbedded-Core, the core system component,
provided by the OpenEmbedded project.
www.allitebooks.com
Meeting the Yocto Project
The Yocto Project is, therefore, a community open source project that aggregates
several companies, communities, projects, and tools, gathering people with the same
purpose to build a Linux-based embedded product; all these components are in the
same boat, being driven by its community needs to work together.
Delineating the Yocto Project
To ease our understanding of the duties and outcomes provided by the Yocto Project,
we can use the analogy of a computing machine. The input is a set of data that
describes what we want, that is, our specification. As an output, we have the desired
Linux-based embedded product.
If the output is a product running a Linux-based operating system, the result
generated is the pieces that compose the operating system, such as the Linux kernel,
bootloader, and the root filesystem (rootfs) bundle, which are properly organized.
To produce the resultant rootfs bundle and other deliverables, the Yocto Project's
tools are present in all intermediary steps. The reuse of previously built utilities
and other software components are maximized while building other applications,
libraries, and any other software components in the right order and with the
desired configuration, including the fetching of the required source code from
their respective repositories such as The Linux Kernel Archives (www.kernel.org),
GitHub, and www.SourceForge.net.
Preparing its own build environment, utilities, and toolchain, the amount of host
software dependency is reduced, but a more important implication is that the
determinism is considerably increased. The utilities, versions, and configuration
options are the same, minimizing the number of host utilities to rely on.
We can list some projects, such as Poky, BitBake, and OpenEmbedded-Core, under
the Yocto Project umbrella, all of them being complimentary and playing specific
roles in the system. We will understand exactly how they work together in this
chapter and throughout the book.
Understanding Poky
Poky is the Yocto Project reference system and is composed of a collection of tools
and metadata. It is platform-independent and performs cross-compiling, using the
BitBake tool, OpenEmbedded Core, and a default set of metadata, as shown in the
following figure. It provides the mechanism to build and combine thousands of
distributed open source projects to form a fully customizable, complete, and coherent
Linux software stack.
[8]
Chapter 1
Poky's main objective is to provide all the features an embedded developer needs.
Poky Build Tool
BitBake Tool (bitbake)
OpenEmbedded-Core (meta)
Yocto-specific Metadata (meta-yocto)
Yocto-specific BSP (meta-yocto-bsp)
Using BitBake
BitBake is a task scheduler that parses Python and Shell Script mixed code. The code
parsed generates and runs tasks, which are basically a set of steps ordered according
to the code's dependencies.
It evaluates all available configuration files and recipe data (known as metadata),
managing dynamic variable expansion, dependencies, and code generation. It keeps
track of all tasks being processed in order to ensure completion, maximizing the use
of processing resources to reduce build time and being predictable. The development
of BitBake is centralized in the bitbake-devel@lists.openembedded.org mailing
list, and its code can be found in the bitbake subdirectory of Poky.
OpenEmbedded-Core
The OpenEmbedded-Core metadata collection provides the engine of the Poky build
tool. It is designed to provide the core features and needs to be as clean as possible.
It provides support for five different processor architectures (ARM, x86, x86-64,
PowerPC, MIPS and MIPS64), supporting only QEMU-emulated machines.
The development is centralized in the openembedded-core@lists.openembedded.
org mailing list, and houses its metadata inside the meta subdirectory of Poky.
[9]
Meeting the Yocto Project
Metadata
The metadata, which is composed of a mix of Python and Shell Script text files,
provides a tremendously flexible system. Poky uses this to extend OpenEmbeddedCore and includes two different layers, which are another metadata subset shown
as follows:
• meta-yocto: This layer provides the default and supported distributions,
visual branding, and metadata tracking information (maintainers, upstream
status, and so on)
• meta-yocto-bsp: This layer, on top of it, provides the hardware reference
boards support for use in Poky
Chapter 7, Diving into BitBake Metadata, explores the metadata in more detail and
serves as a reference when we write our own recipes.
The alliance of OpenEmbedded Project
and Yocto Project
The OpenEmbedded project was created around January 2003 when some core
developers from the OpenZaurus project started to work with the new build system.
The OpenEmbedded build system has been, since its beginning, a tasks scheduler
inspired and based on the Gentoo Portage package system named BitBake. The project
has grown its software collection, and a number of supported machines at a fast pace.
As consequence of uncoordinated development, it is difficult to use OpenEmbedded
in products that demand a more stable and polished code base, which is why Poky
was born. Poky started as a subset of OpenEmbedded and had a more polished
and stable code base across a limited set of architectures. This reduced size allowed
Poky to start to develop highlighting technologies, such as IDE plugins and QEMU
integration, which are still being used today.
Around November 2010, the Yocto Project was announced by the Linux Foundation
to continue this work under a Linux Foundation-sponsored project. The Yocto
Project and OpenEmbedded Project consolidated their efforts on a core build system
called OpenEmbedded-Core, using the best of both Poky and OpenEmbedded,
emphasizing an increased use of additional components, metadata, and subsets.
[ 10 ]
Chapter 1
Summary
This first chapter provided an overview on how the OpenEmbedded Project is
related to the Yocto Project, the components which form Poky, and how it was
created. In the next chapter, we will be introduced to the Poky workflow with steps
to download, configure, and prepare the Poky build environment, and how to have
the very first image built and running using QEMU.
[ 11 ]
Baking Our
Poky-based System
In this chapter, we will understand the basic concepts involved in the Poky
workflow. Let's get our hands dirty with steps to download and configure, prepare
the Poky build environment, and bake something usable. The steps covered here are
commonly used for testing and development. They give us the whole experience of
using Poky and a taste of its capabilities.
Configuring a host system
The process needed to set up our host system depends on the distribution we run on
it. Poky has a set of supported Linux distributions, and if we are new to embedded
Linux development, it is advisable to use one of the supported Linux distributions
to avoid wasting time debugging build issues related to the host system support.
Currently, the supported distributions are the following:
• Ubuntu 12.04 (LTS)
• Ubuntu 13.10
• Ubuntu 14.04 (LTS)
• Fedora release 19 (Schrödinger's Cat)
• Fedora release 20 (Heisenbug)
• CentOS release 6.4
• CentOS release 6.5
• Debian GNU/Linux 7.x (Wheezy)
• openSUSE 12.2
Baking Our Poky-based System
• openSUSE 12.3
• openSUSE 13.1
If our preferred distribution is not in the preceding list, it doesn't mean it is not
possible to use Poky on it. However, it is unknown whether it will work, and we
may get unexpected results.
The packages that need to be installed into the host system vary from one
distribution to another. Throughout this book, you find instructions for Debian and
Fedora, our preferred distributions. You can find the instructions for all supported
distributions in the Yocto Project Reference Manual.
Installing Poky on Debian
To install the needed packages for a headless host system, run the following command:
$: sudo apt-get install gawk wget git-core diffstat unzip texinfo
build-essential chrpath
If our host system has graphics support, run the following command:
$: sudo apt-get install libsdl1.2-dev xterm
The preceding commands are also compatible with the Ubuntu distributions.
Installing Poky on Fedora
To install the needed packages for a headless host system, run the following
command:
$: sudo yum install gawk make wget tar bzip2 gzip python unzip perl
patch diffutils diffstat git cpp gcc gcc-c++ eglibc-devel texinfo
chrpath ccache
If our host system has graphics support, run the following command:
$: sudo yum install SDL-devel xterm
Downloading the Poky source code
After we install the needed packages into our development host system, we
need to get the Poky source code that can be downloaded with Git, using the
following command:
$: git clone git://git.yoctoproject.org/poky --branch daisy
[ 14 ]
Chapter 2
Learn more about Git at http://git-scm.com.
After the download process is complete, we should have the following contents
inside the poky directory:
The examples and code presented in this and the next
chapters use the Yocto Project Version 1.6 and Poky Version
11.0. The code name is Daisy, as reference.
Preparing the build environment
Inside the poky directory, there is a script named oe-init-build-env, which should
be used to set up the build environment. The script must be run as shown:
$: source poky/oe-init-build-env [build-directory]
Here, build-directory is an optional parameter for the name of the directory
where the environment is set; in case it is not given, it defaults to build. The
build-directory is the place where we perform the builds.
It is very convenient to use different build directories. We can work on distinct
projects in parallel or different experimental setups without affecting our
other builds.
[ 15 ]
Baking Our Poky-based System
Throughout the book, we will use build as the build directory. When
we need to point to a file inside the build directory, we will adopt the
same convention, for example, build/conf/local.conf.
Knowing the local.conf file
When we initialize a build environment, it creates a file called build/conf/local.
conf, which is a powerful tool that can configure almost every aspect of the build
process. We can set the machine we are building for, choose the toolchain host
architecture to be used for a custom cross-toolchain, optimize options for maximum
build time reduction, and so on. The comments inside the build/conf/local.conf
file are a very good documentation and reference of possible variables, and their
defaults. The minimal set of variables we probably want to change from the default
is the following:
BB_NUMBER_THREADS ?= "${@oe.utils.cpu_count()}"
PARALLEL_MAKE ?= "-j ${@oe.utils.cpu_count()}"
MACHINE ??= "qemux86"
BB_NUMBER_THREADS and PARALLEL_MAKE should be set to
twice the host processor's number of cores.
The MACHINE variable is where we determine the target machine we wish to build
for. At the time of writing this book, Poky supports the following machines in its
reference Board Support Package (BSP):
•
•
•
•
•
beaglebone: This is BeagleBone
genericx86: This is a generic support for 32-bit x86-based machines
genericx86-64: This is a generic support for 64-bit x86-based machines
mpc8315e-rdb: This is a freescale MPC8315 PowerPC reference platform
edgerouter: This is Edgerouter Lite
The machines are made available by a layer called meta-yocto-bsp. Besides these
machines, OpenEmbedded-Core also provides support for the following:
•
•
•
•
qemuarm: This is the QEMU ARM emulation
qemumips: This is the QEMU MIPS emulation
qemumips64: This is the QEMU MIPS64 emulation
qemuppc: This is the QEMU PowerPC emulation
[ 16 ]
Chapter 2
• qemux86-64: This is the QEMU x86-64 emulation
• qemux86: This is the QEMU x86 emulation
Other machines are supported through extra BSP layers and these are available from
a number of vendors. The process of using an extra BSP layer is shown in Chapter 10,
Exploring External Layers.
The local.conf file is a very convenient way to override several
default configurations over all the Yocto Project's tools. Essentially, we
can change or set any variable, for example, add additional packages to
an image file.
Though it is convenient, it should be considered as a temporary change
as the build/conf/local.conf file is not usually tracked by any
source code management system.
Building a target image
Poky provides several predesigned image recipes that we can use to build our
own binary image. We can check the list of available images running the following
command from the poky directory:
$: ls meta*/recipes*/images/*.bb
All the recipes provide images which are, in essence, a set of unpacked and
configured packages, generating a filesystem that we can use on an actual hardware.
Next, we can see a short description of available images, as follows:
• build-appliance-image: This is a virtual machine image which can be run
by either VMware Player or VMware Workstation that allows to run builds.
• core-image-full-cmdline: This is a console-only image with full support
for the target device hardware.
• core-image-minimal: This is a small image allowing a device to boot, and it
is very useful for kernel and boot loader tests and development.
• core-image-minimal-dev: This image includes all contents of the coreimage-minimal image and adds headers and libraries that we can use in a
host development environment.
• core-image-minimal-initramfs: This core-image-minimal image is used
for minimal RAM-based initial root filesystem (initramfs) and as a part of
the kernel.
[ 17 ]
www.allitebooks.com
Baking Our Poky-based System
• core-image-minimal-mtdutils: This is a core-image-minimal image that
has support for the MTD utilities for use with flash devices.
• core-image-full-cmdline: This is a console-only image with more fullfeatured Linux system functionalities installed.
• core-image-lsb: This is an image that conforms to the Linux Standard Base
(LSB) specification.
• core-image-lsb-dev: This is a core-image-lsb image that is suitable for
development work using the host, since it includes headers and libraries that
we can use in a host development environment.
• core-image-lsb-sdk: This is a core-image-lsb image that includes a
complete standalone SDK. This image is suitable for development using the
target.
• core-image-clutter: This is an image with clutter support that enables
development of rich and animated graphical user interfaces.
• core-image-directfb: This is an image that uses DirectFB instead of X11.
• core-image-weston: This is an image that provides the Wayland protocol
libraries and the reference Weston compositor.
• core-image-x11: This is a very basic X11 image with a terminal.
• qt4e-demo-image: This is an image that launches into the Qt Demo
application for the embedded (not based on X11) version of Qt.
• core-image-rt: This is a core-image-minimal image plus a real-time test
suite and tool appropriate for real-time use.
• core-image-rt-sdk: This is a core-image-rt image that includes a
complete standalone SDK and is suitable for development using the target.
• core-image-sato: This is an image with Sato support and a mobile
environment for mobile devices that use X11; it provides applications such as
a terminal, editor, file manager, media player, and so forth.
• core-image-sato-dev: This is a core-image-sato image that includes
libraries needed to build applications on the device itself, testing and
profiling tools and debugging symbols.
• core-image-sato-sdk: This is a core-image-sato image that includes a
complete standalone SDK and is suitable for development using the target.
• core-image-multilib-example: This is an example image that includes a
lib32 version of Bash, otherwise it is a standard Sato image.
[ 18 ]
Chapter 2
The up-to-date image list can be seen in the Yocto Project Reference Manual.
The process of building an image for a target is very simple. We must run the
following command:
$: bitbake <recipe name>
For example, to build core-image-full-cmdline, run the following command:
$: bitbake core-image-full-cmdline
We will use MACHINE = "qemuarm" in the following examples. It
should be set in build/conf/local.conf accordingly.
Running images in QEMU
As many projects have a small portion that is hardware dependent, the hardware
emulation comes to speed up the development process by enabling sample to run
without involving an actual hardware.
Quick EMUlator (QEMU) is a free and open source software package that performs
hardware virtualization. The QEMU-based machines allow test and development
without real hardware. Currently, the ARM, MIPS, MIPS64, PowerPC, and x86 and
x86-64 emulations are supported.
The runqemu script enables and makes the use of QEMU with the OpenEmbeddedCore supported machines easier. The way to run the script is as follows:
$: runqemu <machine> <zimage> <filesystems>
Here, <machine> is the machine/architecture to be used as qemuarm, qemumips,
qemuppc, qemux86, or qemux86-64. Also, <zimage> is the path to a kernel (for
example, zimage-qemuarm.bin). Finally,<filesystem> is the path to an ext3 image
(for example, filesystem-qemuarm.ext3) or an NFS directory.
[ 19 ]
Baking Our Poky-based System
So, for example, in case we run runqemu qemuarm core-image-full-cmdline, we
can see something as shown in the following screenshot:
We can log in with the root account using an empty password. The system behaves
as a regular system even being used inside the QEMU. The process to deploy an
image in a real hardware varies depending on the type of storage used, bootloader,
and so on. However, the process to generate the image is the same. We explore how
to build and run an image in the Wandboard machine in Chapter 14, Booting Our
Custom Embedded Linux.
Summary
In this chapter, we learned the steps needed to set up Poky and get our first image
built. We ran that image using runqemu, which gave us a good overview of the
available capabilities.
In the next chapter, we will be introduced to Hob, which provides a human friendly
interface for BitBake, and we will use it to build an image and customize it further.
[ 20 ]
Using Hob to Bake an Image
Hob is a human friendly interface for BitBake. It helps us customize images and have
them the way we want. It also enables us to run the image on QEMU after bitbaking
it. It is just like a bakery display; we can pick what we want and use it right away.
Building an image using Hob
Our first step is to set up our build environment, as follows:
$: source poky/oe-init-build-env [build-directory]
We can choose an old build directory or create a new one.
Now, Hob is ready for use. To start it, we should run the following:
$: hob
At startup, Hob performs some parsing tasks, reading the local configuration
and available metadata layers. After a short time, Hob proposes a list of available
machines. We can select, for example, qemuarm.
Once the dependency tree is built, select the desired image, for example, coreimage-full-cmdline.
Using Hob to Bake an Image
The following screenshot shows the MACHINE variable content and the image to be
built in the Hob interface:
With the target MACHINE and image selected, the next step is to choose some
advanced configuration, such as image types (for example, cpio.gz, ext2.bz2,
ext3.gz, jffs2, ubifs, and vmdk) or package formats (rpm, deb, IPK, or TAR). We
can also exclude all packages under the GPLv3 licensing, as shown in Chapter 13,
Achieving GPL Compliance.
From the upper-right hand corner of the window, we can access the two areas
Images and Settings. Images offers access to the built images (from the past),
and Settings performs changes to MACHINE, parallelization, distribution, shared
folders, and BBLAYERS. Hob modifies the build/conf directory contents inside our
build directory. We can use Hob on our already configured build folder, and all
configurations are reflected on Hob. It may be very useful when working on a team.
If we are working to configure the shared environment for a team, we need to pay
attention to the variables DL_DIR and SSTATE_DIR, which are detailed in Chapter 4,
Grasping the BitBake Tool, and Chapter 6, Assimilating Packaging Support.
[ 22 ]
Chapter 3
If we plan to build a standard image, we can click on Build Image and wait for
BitBake to run the required tasks to build it. Otherwise, if we want to change the
recipe set of an image, we can click on Edit image recipe.
Customizing an image with Hob
The following screenshot shows the list of included recipes in the Hob interface:
We can add or remove recipes (there is a search box in the upper-right hand corner)
by selecting or deselecting them. If we click on the recipe name, we can see details
such as its version and license.
[ 23 ]
Using Hob to Bake an Image
From the tabs, we can see the number of selected packages, the list of available
packages, and how the selected packages are grouped, as shown in the
following screenshot:
After clicking on Build packages and waiting for them to be built, we have a second
chance to see the list of selected packages, to know the value of the Estimated image
size, and to decide to remove some application in order to generate a smaller image.
By clicking on the package name, the included files from this package are listed. If a
package is highlighted, its log can be displayed by clicking on Open log.
BitBake resolves all dependencies from the selected packages, including any needed
additional package.
We can wrap the image by clicking on Build image and waiting until our image is
ready, as shown in the following screenshot:
[ 24 ]
Chapter 3
We can start over and change configurations, edit the selected packages, view logs,
or list the files. Or, for images made for the QEMU-based machines, we can click on
Run image and see our image being run inside the QEMU emulator, and the Yocto
Project logo, as shown in the next screenshot:
[ 25 ]
Using Hob to Bake an Image
Hob is a nice tool for image adjustments and addition of few packages on existing
images. It is a great user-friendly interface for simple tasks and may be useful for
teams to make temporary modifications in their images.
Hob is in the process of being replaced by a new tool called Toaster.
At the time of writing this book, Toaster is under heavy development,
and it is still feature incomplete. However, the next Yocto Project
release will supersede Hob, according to the Toaster planned feature
set. So, it is advised to research for Toaster in the Yocto Project
documentation website for more updated information.
Summary
In this chapter, we learned about the different Hob functionalities and configurable
variables. We learned how Hob can be used to easily make changes by teams and
how it takes advantage of a user-friendly user interface for simple tasks.
In the next chapter, we will understand how BitBake does its magic. We will grasp
the parsing, preferences and providers support, dependencies, task handling, and
main task functions.
[ 26 ]
Grasping the BitBake Tool
We now start our journey to understand how the Yocto Project's engine works
behind the scenes. In the preceding chapters, we were introduced to the usual Yocto
Project workflow to create and emulate images.
In this chapter, we will understand the metadata concept and how recipes depend on
each other, and are used by Poky. We will understand how BitBake downloads every
needed source code package and how these packages are stored in the directory used
to build. This chapter also lists the most common tasks used to generate packages,
and determines how packages fit into generated images.
Understanding the BitBake tool
The BitBake task scheduler started as a fork from Portage, which is the package
management system used in the Gentoo distribution. However, nowadays the two
projects have diverged a lot due to the different usage focusses. The Yocto Project
and the OpenEmbedded Project are the most known and intensive users of BitBake,
which remains a separated and independent project with its own development cycle
and mailing list (bitbake-devel@lists.openembedded.org).
As presented in Chapter 1, Meeting the Yocto Project, BitBake is a task scheduler that
parses Python and the shell script mixed code. The code parsed generates and runs
tasks that may have a complex dependency chain, which is scheduled to allow a
parallel execution and maximize the use of computational resources. BitBake can be
understood as a tool similar to GNU Make in some aspects.
In this chapter, we cover the main aspects of BitBake tool. However, for more indepth details about the tool, please refer to the BitBake User Manual.
www.allitebooks.com
Grasping the BitBake Tool
Exploring metadata
The metadata used by BitBake can be classified into three major areas:
• Configuration (the .conf files)
• Classes (the .bbclass files)
• Recipes (the .bb and .bbappend files)
The configuration files define the global content, which is used to provide
information and configure how the recipes will work. One common example of the
configuration file is the machine file that has a list of settings, which describes the
board hardware.
The classes are used by the whole system and can be inherited by recipes, according
to their needs or by default, as in this case with classes used to define the system
behavior and provide the base methods. For example, kernel.bbclass helps the
process of build, install, and package of the Linux kernel, independent of version and
vendor.
The recipes and classes are written in a mix of Python and
shell scripting code.
The classes and recipes describe the tasks to be run and provide the needed
information to allow BitBake to generate the needed task chain. The inheritance
mechanism that permits a recipe to inherit one or more classes is useful to reduce
code duplication, and eases the maintenance. A Linux kernel recipe example is
linux-yocto_3.14.bb, which inherits a set of classes, including kernel.bbclass.
BitBake's most commonly used aspects across all types of metadata (.conf, .bb, and
.bbclass) are shown in the following examples.
The metadata grammar and syntax are detailed in Chapter 7, Diving into
BitBake Metadata.
Parsing metadata
As previously said, there are three metadata groups—configuration, class,
and recipe.
The first parsed metadata in BitBake is configuration metadata, identified by the
.conf file extension. This metadata is global, and therefore, affects all recipes and
tasks which are executed.
[ 28 ]
Chapter 4
BitBake first searches the current working directory for an optional build/conf/
bblayers.conf configuration file, and it is expected to contain a BBLAYERS variable
that is a space-delimited list of layer directories. For each directory in this list, a
build/conf/layer.conf file is searched for and parsed with the LAYERDIR variable
being set to the directory where the layer was found. This process automatically sets
up BBPATH and other variables for a given build directory for the user.
The order of the listed layers in the BBLAYERS variable is
followed by BitBake when parsing the metadata. In case
your layer needs to be parsed first, be sure to have it listed
in the right order in BBLAYERS.
BitBake then expects to find meta/conf/bitbake.conf in the user-specified BBPATH.
The configuration file generally has the include directives to pull in many other
metadata such as architecture specific ones, machine configuration files, and the
build/conf/local.conf file. Only variable definitions and the include directives are
allowed in configuration files (.conf). BitBake's classes (.bbclass), as said earlier, are
its rudimentary inheritance mechanism. They're parsed when an inherit directive is
encountered, and they are located in classes/, relative to the directories in BBPATH.
A BitBake recipe (.bb) is a logical unit of tasks to be executed. Normally, this is a
package to be built. Inter-recipe dependencies are obeyed. The files themselves are
found via the BBFILES variable, which is set to a space separated list of the .bb files,
and handles wildcards.
Dependencies
In order to accomplish the dependency, the recipes must declare what they need
to have available during the build process. BitBake ensures that the build-time
dependencies are satisfied before starting the recipe build. This is easier to understand
if we think about an application that uses a library. So, this library must be built and its
headers made available for use, before the application itself can be built. The DEPENDS
variable is used in a recipe to inform BitBake about the build-time dependency.
When an application depends on something to run, this is called a runtime
dependency. This is common for shared data among applications (for example,
icons), which is used only when running the application but not used during its
build process or when an application calls another during its execution. The runtime
dependencies can be expressed using the RDEPENDS variable in a recipe.
[ 29 ]
Grasping the BitBake Tool
With the recipe dependencies chain, BitBake can sort all the recipes in a feasible
order for the build. This permits BitBake to organize tasks in the following ways:
• Recipes that do not have a dependency relation are built in parallel
• Dependent recipes are built in a serial order, sorted in a way the
dependencies are satisfied
Every recipe included in the runtime dependencies is put in the build
list. This sounds obvious, but even though they are not used during
the build, they are included because they need to be ready for use so
that the resulting binary packages are installable.
Preferring and providing recipes
The dependency relation between recipes is core to BitBake and the build tool as a
whole. It is defined inside each recipe file, with a variable which describes on what
a recipe depends (DEPENDS) and what a recipe provides to the system (PROVIDES).
These two variables together build the dependency graph used by BitBake during
the dependency resolution.
So, if a recipe foo_1.0.bb depends on bar, BitBake lists all recipes providing bar.
The bar dependency can be satisfied by the following:
• The bar_<version>.bb format as every recipe provides itself
• A recipe with the PROVIDES variable set to bar
A dependency can be satisfied by several recipes (for example, two or more recipes
have PROVIDES += "bar"). In this case, we must inform BitBake the provider to use.
The virtual/kernel provider is a clear example where this mechanism is used. The
virtual/ namespace is the convention adopted when we have a set of commonly
overridden providers.
All recipes that require the kernel to build can add virtual/kernel to the dependency
list (DEPENDS), and BitBake makes sure to satisfy the dependency. When we have
more than one recipe with an alternative provider, we must choose one to be used, for
example, PREFERRED_PROVIDER_virtual/kernel = "linux-mymachine".
The virtual/kernel provider is commonly set in the machine definition file as it
may vary from one machine to another. We see how to create a machine definition
file in Chapter 11, Creating Custom Layers.
[ 30 ]
Chapter 4
When BitBake cannot satisfy a dependency, due to a missing
provider for it, an error is raised.
The preference cannot be set per recipe; so, in the same system, two recipes cannot
use different providers for the same dependency.
When BitBake has two providers with different versions, it uses the higher version
by default. We can force BitBake to use a different version by using PREFERRED_
VERSION. This is commonly found in BSPs, such as bootloaders, where vendors use
different versions depending on the board.
When we have a development or an unreliable version of a recipe, and we do not
want it to be used by default, we can use DEFAULT_PREFERENCE = "-1" in the recipe
file. So, even if the version is greater, it is not taken without it being explicitly set
(using PREFERRED_VERSION for it).
Fetching the source code
One of the main features supported by BitBake is source code fetching. This support
has been designed to be as modular and as flexible as possible. The mechanism used
by BitBake to fetch the source code is internally called as fetcher backend. There
are several fetcher backends supported, which can be configured to align the user
requirements and optimize source code fetching.
When the Poky source code is downloaded, what is actually copied is the metadata
and the BitBake tool. All other source code is fetched on demand. Every Linuxbased system includes the Linux kernel and several other utilities that form the
root filesystem, such as openssh. The OpenSSH source code is available from its
upstream website as a tar.gz file hosted on a FTP server; the Linux kernel release
may be fetched from a Git repository.
BitBake offers support for many different fetcher modules that allow retrieval of
tarball files and a number of other protocols such as Git, Subversion, Bazaar, OSC,
HTTP, HTTPS, FTP, CVS, Mercurial, Perforce, or SSH.
Remote file downloads
BitBake supports several methods for remote file downloads. The most commonly
used are http://, https://, and ftp://. We won't cover the internal details of how
BitBake handles it but focus on the visible effects of it.
[ 31 ]
Grasping the BitBake Tool
When BitBake executes the do_fetch task in a recipe, it checks the SRC_URI
contents. If we look at, for example, the libjson recipe (available at meta/recipesdevtools/libjson/libjson_0.9.bb), the processed variables are the following:
SRC_URI = "http://oss.metaparadigm.com/json-c/json-c-${PV}.tar.gz"
SRC_URI[md5sum] = "3a13d26...de24abae"
SRC_URI[sha256sum] = "702a486...39439475"
BitBake expands the PV variable to the package version, 0.9 in this example,
downloads the file from the URL http://oss.metaparadigm.com/json-c/jsonc-0.9.tar.gz, and then saves it at the path pointed out by the DL_DIR variable.
After the download is complete, BitBake compares the md5sum and sha256sum values
of the downloaded file with the values from recipe, and if both values match, it
creates a ${DL_DIR}/json-c-0.9.tar.gz.done file to mark the file as successfully
downloaded and checked. This way, when BitBake looks for the file next time, it
knows that it can safely be reused, and it skips the file's checksum verification. This
process happens for every remote file that BitBake downloads.
By default, the DL_DIR variable points to build/downloads. You can
override this using the build/conf/local.conf file as:
DL_DIR = "/my/download-cache"
This makes it easy to share the same download cache among several
different build directories, thus saving time and bandwidth.
Git repositories
One of the most commonly used source control management systems in use is Git.
BitBake has a solid support for it, and the Git backend is used when the do_fetch
task is run and finds a git:// URL at the SRC_URI variable.
The default way for the BitBake's Git backend to handle the repositories is to clone
the repository into ${DL_DIR}/git2/<git URL>. For example, check the following
quote from the linux-firmware_git.bb recipe found in meta/recipes-kernel/
linux-firmware/linux-firmware_git.bb inside Poky:
SRCREV = "600caefd83a406540b2a789be6415e44c9b87add"
...
SRC_URI =
"git://git.kernel.org/pub/scm/linux/kernel/git/firmware/linuxfirmware.git"
[ 32 ]
Chapter 4
Here, the linux-firmware.git repository is cloned into ${DL_DIR}/git2/git.
kernel.org.pub.scm.linux.kernel.git.firmware.linux-firmware.git.
This directory name is chosen to avoid conflicts between other possible Git
repositories with the same project name. The SRCREV variable is used by the do_
fetch task to ensure the repository has the needed Git revision and forces an update
in case it does not; it's used by the do_unpack task to set up the work directory in the
wanted source revision.
When the SRCREV variable points to a hash not available in the master
branch, we need to use the branch=<branch name> parameter,
as follows:
SRC_URI = "git://myserver/myrepo.git;branch=mybranch"
In the cases when the hash used points to a tag, which is not available in
a branch, we need to use the nobranch=1 option as the following:
SRC_URI = "git://myserver/myrepo.git;nobranch=1"
Other repositories
The remote file and the Git repository are the most commonly used fetch backends
of BitBake. The other source code management supports vary in the implementation,
but the general ideas and concepts are the same.
Optimizing the source code download
To improve the robustness of source code download, Poky provides the mirror
mechanism that can be configured in order to do the following:
• Provide a centrally preferred server for download
• Provide fallback servers
In order to provide this robust download mechanism, BitBake follows some steps.
During the build, the first BitBake step is to search for the source code within the
local download directory (pointed by DL_DIR). In case of failure, the next step is to
try locations defined by the PREMIRRORS variable. In a recurrent case of failure, it
searches the locations pointed out in the MIRRORS variable.
[ 33 ]
Grasping the BitBake Tool
The following sections explain the PREMIRRORS and MIRRORS variables:
• PREMIRRORS: The second step in the download mechanism performed by
BitBake is controlled by the variable PREMIRROR. For example, the Poky
distribution sets it as the following:
PREMIRRORS_prepend = "\
git://.*/.* http://www.yoctoproject.org/sources/ \n \
ftp://.*/.* http://www.yoctoproject.org/sources/ \n \
http://.*/.* http://www.yoctoproject.org/sources/ \n \
https://.*/.* http://www.yoctoproject.org/sources/ \n"
The preceding code prepends the PREMIRROR variable to change and instruct
the build system to intercept any Git, FTP, HTTP, and HTTPS requests, and
it redirects them to the http://www.yoctoproject.org/sources/ sources
mirror.
In case the desired component is not available in the source mirror, BitBake
falls back to the MIRRORS variable.
• MIRRORS: The variable MIRRORS provides a set of alternative URL addresses
for some servers where the source component may be found. BitBake tries
one after another, and if all available mirrors fail, it raises an error.
We can take advantage of this download mechanism to save the download time by
providing a local pull directory or a local network server by adding the following
code to build/conf/local.conf:
SOURCE_MIRROR_URL ?= "file:///home/you/your-download-dir/"
INHERIT += "own-mirrors"
Here, SOURCE_MIRROR_URL can point to a local directory or to any server URL with
the supported fetcher backend.
In case the goal is to have a shareable download cache, it is advisable to
enable the tarball generation for the SCM backends (for example, Git) in
the download folder with BB_GENERATE_MIRROR_TARBALLS = "1"
in build/conf/local.conf.
[ 34 ]
Chapter 4
Disabling network access
Sometimes, we need to ensure that we don't connect to the Internet during the build
process. There are several valid reasons for this, such as the following:
• Policy: Our company does not allow external source to be included in a
product without a proper legal validation and review
• Network cost: When we are on the road using a mobile broadband, the cost
of data may be too expensive as the data to download may be big
• Build speed: When we are sure we have already downloaded the required
source code, we can turn off the network access in order to speed up the
building process. This setup is commonly used in autobuild servers
• Lack of network access: Sometimes, we do not have access to a network
In order to disable the network connection, we need to add the following code in the
build/conf/local.conf file:
BB_NO_NETWORK = "1"
Understanding BitBake's tasks
BitBake uses execution units, which are in essence a set of clustered instructions
that run in sequence. These units are known as tasks. There are many tasks being
scheduled, executed, and checked by BitBake during every recipe build, provided by
classes to form the framework that we use to build a recipe. Some are important to
be understood as we often use, extend, implement, or replace them ourselves when
writing a recipe.
Run the following command:
$: bitbake <recipe>
BitBake runs a set of scheduled tasks. When we wish to run a specific task, we can
use the following command:
$: bitbake <recipe> -c <task>
To list the tasks defined for a recipe, we can use the following command:
$: bitbake <recipe> -c listtasks
[ 35 ]
Grasping the BitBake Tool
We will briefly describe each of these here:
• do_fetch: The first step when building a recipe is fetching the needed
source. This is done using the fetching backends feature we discussed
previously in this chapter. It is important to point out that fetching a source
or a file does not mean it is a remote source. In fact, every file needed during
the recipe build must be fetched so that it is made available in the WORKDIR
directory. We will learn more about the build directory and its contents in
Chapter 5, Detailing the Temporary Build Directory.
All downloaded content is stored in the download folder (the DL_DIR
variable), so all external source code is cached to avoid redownloading it
every time we need the same source.
• do_unpack: The natural subsequent task after the do_fetch task is do_unpack.
It is responsible for unpacking source code or to check out the requested
revision or branch, in case the referenced source uses a SCM system.
• do_patch: Once the source code has been properly unpacked, BitBake initiates
the process of adapting the source code. This is done by the do_patch task.
Every file fetched by do_fetch, with the .patch extension, is assumed to be a
patch to be applied. This task applies the list of patches needed.
The process of applying a patch uses the S variable, which points to the source
code. The default value used for S is ${WORKDIR}/${PN}-${PV}, and it is used
for the do_patch, do_configure, do_compile and do_install tasks.
• do_configure, do_compile, and do_install: The tasks do_configure,
do_compile, and do_install are performed in this order. Some recipes
may omit one task or another. It is important to note that the environment
variables defined in the tasks are different from one task to another.
The tasks vary a lot from one recipe to another. Poky provides a rich
collection of predefined tasks in the classes, which ought to be used when
possible. For example, when the Autotools class is inherited by a recipe, it
provides a known implementation for the do_configure, do_compile and
do_install tasks.
• do_package: The do_package task splits the files installed by the recipe
into logical components such as debugging symbols, documentation, and
libraries. The do_package task ensures that files are split up and packaged
correctly. We will cover the packaging details in more depth in Chapter 6,
Assimilating Packaging Support.
[ 36 ]
Chapter 4
Extending tasks
When the task content does not satisfy our requirements, we replace it (providing
our own implementation) or append it. As we learn more extensively when learning
about the BitBake metadata syntax in Chapter 7, Diving into BitBake Metadata, the _
append and _prepend operators can be used to extend a task with extra content. The
new content is concatenated in the original task. For example, to extend a
do_install task, we can use the following code:
do_install_append() {
# Do my commands
}
The mechanism using which we can extend existing recipes is covered in Chapter 12,
Customizing Existing Recipes.
Generating a root filesystem image
One of the most common uses of Poky is the rootfs image generation (rootfs).
The rootfs image should be seen as a ready-to-use root filesystem for a target. The
image can be made up of one filesystem or may include other artifacts to be available
during its generation as the Linux kernel, device tree, and bootloader binaries and
other filesystems. The process to generate the image is composed of several steps,
and its most common usages are the following:
1. Generate the rootfs directory.
2. Create the required files.
3. Wrap the final filesystem accordingly to the specific requirements (it may be
a disk file with several partitions and contents).
4. Finally, compress it, if applicable.
All these steps are performed by subtasks of do_rootfs.
The rootfs is basically a directory with the desired packages installed (the package
generation is covered in Chapter 6, Assimilating Packaging Support), with the needed
tweaks applied just afterwards. The tweaks make minor adjustments applied in
the rootfs contents; for example, when building a development image, rootfs is
adjusted to allow us to log in with root without a password.
[ 37 ]
www.allitebooks.com
Grasping the BitBake Tool
The list of packages to be installed into rootfs is defined by a union of packages
listed by IMAGE_INSTALL and the packages included by IMAGE_FEATURES; the image
customization is detailed in Chapter 11, Creating Custom Layers. Each image feature
can include extra packages for installation, for example, dev-pkgs that installs
development libraries and headers of all packages listed to be installed in rootfs.
The list of packages to be installed is now filtered by the variable PACKAGE_EXCLUDE,
which lists the packages that should not be installed. The packages listed in PACKAGE_
EXCLUDE are only excluded from the list of packages to be explicitly installed.
Packages listed in PACKAGE_EXCLUDE are installed into rootfs
if it is needed to satisfy a runtime dependency.
Having the final set of packages to install, the do_rootfs task can initiate the process
of unpacking and configuring each package, and its required dependencies, into
the rootfs directory. This is done using a package backend's (DEB, IPK, or RPM)
specific set of subtasks as it actually uses the package management system using
a local package feed to do this step. The package feed is explained in Chapter 6,
Assimilating Packaging Support.
With the rootfs contents unpacked, the post installation scripts of the referred
packages must run to avoid the penalty of running them during first boot. Some scripts
may need to be run in the target to succeed, and there is no problem with that, except
in the case we use the read-only-rootfs image feature. The post-installation script
and the other variants of it are covered in Chapter 6, Assimilating Packaging Support.
All the post-installation scripts must be successful in case of a
read-only rootfs directory.
The rootfs optimization is then executed. A prelink process optimizes the dynamic
linking of shared libraries to reduce startup time of executables; the mklibs process
optimizes the size of the libraries removing the unused symbols.
Now, the directory is ready to generate the filesystem. IMAGE_FSTYPES lists the
filesystem to be generated, for example, EXT3 or UBIFS.
After do_rootfs has completely finished, the generated image file is
placed in <build-dir>/tmp/deploy/image/<machine>/.
[ 38 ]
Chapter 4
The process to create our image and the possible values for IMAGE_FEATURES and
IMAGE_FSTYPES are described in Chapter 11, Creating Custom Layers.
Summary
In this chapter, we learned the metadata concept, how recipes depend on each other,
and how Poky deals with dependencies. We also got a better view of the behind-thescenes tasks done by BitBake to download all the required source code, to store it in
the build directory used to build and generate packages, as well as and how these
packages fit into generated images.
In the next chapter, we will see the contents of the build directory after complete
image generation and how BitBake uses it in the baking process, including the
contents of the temporary build directory and its generated files.
[ 39 ]
Detailing the Temporary
Build Directory
In this chapter, we will understand the contents of the temporary build directory
after a complete image generation and see how BitBake uses it in the baking process.
We will learn how some of these directories can assist our understanding when
things do not work as expected, providing a valuable source of information.
Detailing the build directory
The build directory is a central information and artifact source for every Yocto
Project's tool user. Its main directories are the following:
• conf: This has the configuration files we use to control Poky and BitBake.
The first use of this directory was in Chapter 2, Baking Our Poky-based System.
It stores configuration files such as build/conf/local.conf and build/
conf/bblayers.conf.
• downloads: This stores the downloaded tarballs. It can be seen as the
download cache and has been detailed in Chapter 4, Grasping the BitBake Tool.
• sstate-cache: This has the packaged data snapshots. It is a cache mainly
used to speed up the build process. This folder is detailed in Chapter 6,
Assimilating Packaging Support.
• tmp: This is a temporary build directory. It is very handy; we will understand
it better in the following sections.
Detailing the Temporary Build Directory
Constructing the build directory
The Yocto Project's tool inputs and outputs were already detailed in a high abstract
level in the previous chapters. We already know that BitBake uses metadata to
generate a set of artifacts, including images. Besides the generated artifacts, BitBake
creates much more content during the process, which may be used in several ways,
dependent upon our goals.
During the build process, BitBake performs several tasks and modifies the build
directory. We can understand it better following the usual BitBake flow, as follows:
• Fetching: The first action realized by BitBake is to download the source code.
This step modifies the build directory, including a new tarball inside the
build/download directory.
• Source preparation: After the fetching of the source code is complete, it
must be prepared for use. This may involve, for example, the unpacking of a
tarball or a clone of a local cached Git directory (from the download cache).
The source code is prepared in the build/tmp/work directory. When the
source code is ready, the needed modifications are applied (for example,
applying needed patches).
• Configuration and building: With the ready-to-use source code, the building
process can start. It involves the configuration of build options (for example,
./configure) and building (for example, make).
• Installing: The built artifacts are then installed (for example, make install)
in a proper directory under build/tmp/work/<...>/image.
• Wrapping the sysroot: The libraries, headers, and other files that need to be
shared for cross-compilation are copied (and sometimes modified) in build/
tmp/sysroot.
• Creating the package: The packages are generated using the installed
contents, splitting the files into subpackages and creating the packages (for
example, RPM).
Exploring the temporary build directory
It is critical to understand the temporary build directory (build/tmp). The
temporary build directory is created just after the start of a build, and it's especially
important to help us to identify why something didn't behave as expected.
[ 42 ]
Chapter 5
The contents of the build/tmp directory are shown in the following figure:
build/tmp/
abi_version
buildstats
cache
deploy
log
saved_tmpdir
sstate-control
stamps
sysroots
work
work-shared
The most important directories found within it are the following:
• deploy: This contains the build products such as images, packages,
and SDKs
• sysroots: This contains the shared libraries, headers, and utilities that are
used in the process of building recipes
• work: This contains the working source code, a task's configuration, execution
logs, and the contents of generated packages.
Understanding the work directory
The build/tmp/work directory is split by architecture. For example, when working
with the machine qemuarm, we have the following four directories:
• all-poky-linux
• armv5te-poky-linux-gnueabi
• qemuarm-poky-linux-gnueabi
• x86_64-linux
[ 43 ]
Detailing the Temporary Build Directory
The directories found here and their contents are architecture and machine
dependent. We shouldn't take this as a final list, only as an illustration. The directory
x86_64-linux is used to build the host sysroot content, which is detailed in the
next section. The directory all-poky-linux holds the working build directories for
the packages that are architecture agnostic. This fragmented structure is necessary
to allow building multiple machines and architectures within one build directory
without conflicting with each other.
The target machine we use is qemuarm. This machine is an emulation of the ARM
Versatile Platform Baseboard with the ARM926EJ-S CPU emulation that supports
the ARMv5TE instructions. Poky treats qemuarm as a type of ARMv5TE because
some hardware features, may not be available in one device or another, even when
they are supported by the CPU. Machine-specific recipes are built in the machine
directory (qemuarm-poky-linux-gnueabi in this case) while the architecture-specific
packages are built in the architecture-specific directory (armv5te-poky-linuxgnueabi in this case).
The build/tmp/work work directory is very useful when checking misbehavior or
build failures. Its contents are stored in subdirectories following the pattern:
<arch>/<recipe name>/<software version>
Some of the directories under this tree are:
• <sources>: This is an extracted source code of the software to be built. This
directory is pointed by the WORKDIR variable.
• image: This contains the files installed by the recipe (pointed by the D
variable).
• packages: The extracted content of packages are stored here.
• packages-split: The content of packages, extracted and split, are stored
here. This has a subdirectory for each package.
• temp: This stores BitBake's task code and execution logs.
The most commonly checked subdirectories are under the sysroots directory,
which provides the artifacts used during cross-compilation as compilers, utilities,
and libraries for the host and target; they are also checked under the build/tmp/
work directory, which holds the working build directory. These directories provide
valuable information for debugging.
In order to reduce disk usage, we can automatically remove the work
directory after each recipe compilation cycle, adding INHERIT +=
"rm_work" in the build/conf/local.conf file.
[ 44 ]
Chapter 5
The structure of the work directory is the same for all architectures. For every recipe, a
directory with the recipe name is created. Taking the machine-specific work directory
and using the sysvinit-inittab recipe as an example, we see the following:
work/qemuarm-poky-linux-gnueabi/sysvinit-inittab/2.88dsf-r10/
deploy-rpms
image
inittab
license-destdir
package
package-split
pkgdata
pseudo
sysroots-destdir
sysvinit-2.88dsf
sysvinit-inittab.spec
temp
The sysvinit-inittab recipe is a good example, which even without a machinespecific object code is a machine-specific one. It contains the inittab file that defines,
among other things, the serial console devices to spawn the login process, and
this varies from machine to machine as the UART device depends on the machine
schematic layout.
The directories shown in the preceding figure that are not detailed here
are used by the build system.
You should not need to work with them, except if you are working on
build tool development.
The work directory is very useful for debugging purposes; we cover this in Chapter 9,
Debugging with the Yocto Project.
[ 45 ]
Detailing the Temporary Build Directory
Understanding the sysroot directories
The content of the sysroots directory is shown in the following figure:
sysroots /
qemuarm
qemuarm-tcbootstrap
x86_64-linux
We can see three directories: qemuarm, qemuarm-tcbootstrap, and x86_64-linux.
The qemuarm-tcbootstrap directory is used as an intermediate directory for the
compiler bootstrap as it needs some artifacts (for example, the libc headers), which
is only useful for debugging if working on toolchain development.
The x86_64-linux directory is used (the name depends on the host machine
architecture, for example, i686) to build host tools and libraries. The host sysroot is
critical of the cross-compilation process as it provides the utilities used during the
whole build process as pkg-config, autotool, make, git, and every other utility that
is called during the build process; the binaries are built using the headers and libraries
of the directory in order to isolate it completely from the GNU/Linux host distribution
running in our development machine, and it is known as the host sysroot.
Our target machine (in our example, qemuarm) uses the qemuarm (target is sysroot)
directory for headers and libraries required during the build. The utilities that must
be run as part of the build process are called from the host sysroot so that both the
sysroot directories play a core role in the cross-compiling process.
One common mistake to avoid, when designing a library recipe, is not getting
its contents installed properly (the headers, static, and shared libraries) into the
directory pointed by variable D so that Poky can, most of time, make the right
installation of the needed files to sysroot. When we see a missing header or a link
failure, we must double-check if our sysroot (target and host) contents are correct.
[ 46 ]
Chapter 5
Summary
In this chapter, we explored the contents of the temporary build directory after a
complete image generation, and saw how BitBake uses it along the baking process.
We then learned how to use these directories for debugging.
In the next chapter, we will get a better understanding of how packaging is done in
Poky, how to use package feeds and the PR service, and how they may help in our
product maintenance.
[ 47 ]
www.allitebooks.com
Assimilating
Packaging Support
This chapter presents the key concepts to understand Poky and BitBake; it teaches
us the concepts related to packaging. We will learn about the supported package
formats, what shared state cache is and why it is needed, the package versioning
components, how to set up and use package feeds to support our development
process, and more.
Using supported package formats
The main goal of most of Poky's tasks is to produce packages. The other tasks use
built packages to make other deliverables such as images and toolchains.
A recipe may generate one or more packages as a result of it being executed by
BitBake; on the other hand, images and toolchains are made of several packages that
are unpacked and configured to accomplish the intended goal. The generated result
is wrapped in such a way that it can be installed into one or more images, or it can be
deployed for later use.
List of supported package formats
Currently, BitBake supports four different package formats:
• RPM: Originally named Red Hat Package Manager, but now known as the
RPM package format since its adoption to several other Linux distributions,
it is used by several popular Linux distributions such as SuSE, OpenSuSE,
Red Hat, Fedora, CentOS, among others.
Assimilating Packaging Support
• DEB: The Debian Package Manager is used by Debian and several other
Debian-based distributions; the most widely known among them are Ubuntu
Linux and Linux Mint.
• IPK: The Itsy Package Management System was a lightweight package
management system designed for embedded devices that resembled Debian's
package format. The initial project was discontinued, and the several
embedded build systems and distributions that used it were moved to the
Opkg fork made by OpenMoko; it is also used by the OpenWrt project.
Currently, the OpenEmbedded-Core, and as a consequence Poky, uses the
Opkg package manager to support the IPK format.
• TAR: This is derived from the tape archive .tar, and it is a widely used
tarball file type used to group several files into only a single file.
Choosing a package format
The support for the formats is provided using a set of classes (package_rpm,
package_deb, and package_ipk). We can select one or more formats using the
PACKAGE_CLASSES variable, as shown in the following example:
PACKAGE_CLASSES ?= "package_rpm package_deb package_ipk"
This can be done, for example, in the build/conf/local.conf file. Using the
PACKAGE_CLASSES variable, we can generate the packages in one or more formats.
Images are created from the first package format found in
PACKAGE_CLASSES.
Poky defaults to the RPM package format, which uses the smart package manager.
However, the most adequate format depends on several aspects, including package
format specific features, memory and resource usage, and so on. Another aspect is
habit; for example, OpenEmbedded-Core users often feel more comfortable using
IPK and opkg as the package manager as it is OpenEmbedded-Core's default and
offers a smaller footprint in memory and resource usage. On the other hand, people
used to Debian-based systems may prefer to use the APT and DEB package format for
their products.
[ 50 ]
Chapter 6
Running code during package installation
Packages can use scripts as part of their installation and removal process. The
included scripts are defined as follows:
• preinst: This executes before the package is unpacked. Services should
be stopped during the execution of preinst to permit the installation or
upgrade.
• postinst: This typically completes any required configuration of the
package after it has been unpacked. Many postinst scripts then execute any
command necessary to start or restart a service once a new package has been
installed or upgraded.
• prerm: This usually stops any daemon that is associated with a package. It is
executed before the removal of files associated with the package.
• postrm: This commonly modifies links or other files created by the package.
The scripts are supposed to be run after the package installation (postinst) is run
during the root filesystem creation. If the script returns a success value, the package
is marked as installed. If the script execution returns an error, the package is marked
as unpacked. All packages marked as unpacked have their scripts executed again
immediately in the first boot of the image.
In order to add a postinst script, we can use the following code as example:
pkg_postinst_${PN} () {
#!/bin/sh -e
# Insert commands above
}
Instead of using the package name itself, we can use the variable PN that
automatically expands the package name of the recipe.
To delay the script execution so that it runs on the target device itself (we should
avoid this), we use the following structure in the post-installation script:
pkg_postinst_${PN} () {
#!/bin/sh -e
if [ x"$D" = "x" ]; then
# Insert commands here
else
exit 1
fi
}
[ 51 ]
Assimilating Packaging Support
In the previous example, we can see how the execution of the script is delayed.
If the variable $D has a value, the script returns an error and the package is set as
unpacked. It means that any command inserted in the conditional if section is
executed only if the variable $D is unset.
We can also skip postscript execution, at the rootfs creation time, for example, to
avoid trying to start a daemon at that time, but yet ensure that it is properly started
when being upgraded in the device. Here is one example:
pkg_postinst_${PN} () {
#!/bin/sh -e
if [ x"$D" != "x" ]; then
exit 0
fi
# Insert commands here to restart
}
When we generate an image with read-only-rootfs in IMAGE_FEATURES, all postinstallation scripts must succeed. If any script returns an error and the package is
set as unpacked only, forcing the script to be run after the root filesystem is created,
the do_rootfs task fails. This check during build time ensures that we identify the
problem while building the image rather than during the initial boot operation in the
target device due to the impossibility of writing the filesystem.
Make sure all the pkg_postinst script execution for installed
packages is feasible during do_rootfs. This is required in case readonly-rootfs is in IMAGE_FEATURES.
It is important to highlight that one of the most common mistakes when creating
post-installation scripts is the lack of the variable $D in front of absolute paths. This
allows two possible uses of $D, ensuring paths are valid in both the host and target
environments and checking to determine which environment is being used as a
method to take appropriate actions.
Another common mistake is the attempt to run processes that are specific to or
dependent on the target architecture. The easiest solution, in this case, is to postpone
the script execution of the target.
[ 52 ]
Chapter 6
Understanding shared state cache
The default behavior of Poky is to build everything from scratch unless BitBake
determines that a recipe does not need to be rebuilt. The main advantage of building
everything from scratch is that the final result is fresh and there is no risk of previous
data causing problems. However, rebuilding everything requires computational time
and resources.
The strategy to determine whether a recipe must be rebuilt is complex. Basically,
BitBake tries to track as much information as possible about every task, variable, and
code used in the build process. BitBake then generates a checksum for all involved
information for every task.
Poky uses all this information provided by BitBake to store snapshots of those tasks
as a set of packaged data generating a cache, which is called shared state cache
(sstate-cache). This cache wraps the contents of each task output in packages
stored in the SSTATE_DIR directory. Whenever BitBake prepares to run a task, it
first checks the existence of a sstate-cache package that matches. If the package is
present, BitBake uses the prebuild built package.
This is a very simple view of the whole shared state mechanism, which is quite a
complex piece of code. For an advanced overview, it is advised to read the Shared
State Cache section of the Yocto Project Reference Manual.
When using Poky for several builds, we must bear in mind that sstate-cache needs
to be cleaned from time to time since it keeps growing as more and more cached data
is added for every build. There is an easy way of cleaning it, as follows:
$: ./scripts/sstate-cache-management.sh --remove-duplicated -d -cache-dir=<path to sstate-cached>
This removes the duplicated and old data from the cache.
When we need to rebuild from scratch, we either remove the build/
tmp so that we can use sstate-cache to speed up the build or we
remove both build/tmp and sstate-cache so that no cache is
reused during the build.
[ 53 ]
Assimilating Packaging Support
Explaining package versioning
Package versioning is used to differentiate the same package in different stages
of its lifecycle. From Poky's perspective, it is also used as part of the equation that
generates the checksum used by BitBake to verify whether a task must be rebuilt.
The package version, also known as PV, plays a central role when we select which
recipe to build. The default behavior of Poky is to always prefer the newest recipe
version, unless there is an explicit different preference, as we discussed in Chapter
4, Grasping the BitBake Tool. For example, consider that we have two versions of the
recipe myrecipe—myrecipe_1.0.bb and myrecipe_1.1.bb. BitBake, by default,
builds the recipe with version 1.1.
Inside the recipe, we may have other variables that compose package versioning
with the PV variable. These are package epoch, known as PE and package revision,
known as PR.
The PE variable has a default value of zero and is used when the package version
schema is changed, breaking the possibility of usual ordering. The epoch is
prepended in the package version, forcing a higher number when needed. For
example, if a package uses the date to compose the PV variables such as 20140101
and 20140201, the version schema is changed for a reason, and a new version 1.0
is released, it is impossible to determine whether version 1.0 is higher than version
20140201. So, PE = "1" is used, forcing the version 1.0 to be higher than 20140201
since 1:1.0 is greater than 0:20140101.
The PR variable has a default value of r0 and is used as part of package versioning.
When it is updated, it forces BitBake to rebuild all tasks of a specific recipe. We can
update it manually in the recipe metadata to force a rebuild we know is needed.
Although the approach of manually setting the PR variables inside recipes may seem
attractive, it is very fragile because it relies on human interaction and knowledge
when it is required. Since Yocto Project 1.5, called Dora, BitBake uses the task
checksums to control what needs to be rebuilt, and the manual PR increment is only
used in extremely rare cases when the task checksum does not change.
In the next section, we will get a better understanding of how PR are handled
nowadays, using the PR service.
[ 54 ]
Chapter 6
Package feeds
As we discussed in Chapter 4, Grasping the BitBake Tool, the packages play a central
role as images and SDKs rely on them. In fact, do_rootfs uses a local package
repository to fetch. This repository is known as package feed.
There is no reason for this repository to be used just for the images or SDK build
steps. In fact, there are several valid reasons to make this repository remotely
accessible internally in our development environment or publically for use in the
field. Some of these reasons are listed, including the following:
• Easy testing of an updated application during development stage, without
the need of a complete system reinstallation
• Make additional packages more flexible so that they can be installed in a
running image
• Update products in the field
To produce a solid package feed, we must ensure that we have consistent increments
in the package revision every time the package is changed. It is almost impossible
to do this manually, and the Yocto Project has a service, the PR service, specially
designed to help in this area.
PR service, which is part of BitBake, is used in order to increment the PR without
human interaction every time BitBake detects a checksum change in a task.
Essentially, it injects a suffix in PR in the format ${PR}.X. For example, considering
PR = "r34" after subsequent PR service interactions, the PR value becomes r34.1,
r34.2, r34.3, and so on.
When using Poky to generate images and not targeting a remote package feed, PR
service is not required because BitBake triggers the needed rebuilds due to the task
checksum changes. The use of PR service is critical for solid package feeds, as it
requires the version increase in a linear fashion.
Even though we ought to use PR service to have a solid package
versioning, it does not exclude the need of setting PR manually in
exceptional cases.
By default, PR service is not enabled or running. To enable it to run locally, we must
set the PRSERV_HOST variable in the BitBake configuration, for example, in build/
conf/local.conf as the following:
PRSERV_HOST = "localhost:0"
[ 55 ]
Assimilating Packaging Support
This approach is adequate when the build happens in a single computer, which
builds every package of the package feed. BitBake starts and stops the server at each
build and increases the needed PR values automatically.
For a more complex setup, with multiple computers working against a common,
shared package feed, we must have a single PR service running, which is used by all
building systems associated with the package feed. In this case, we need to start PR
service in the server using the bitbake-prserv command, as shown:
$: bitbake-prserv ‐‐host <ip> ‐‐port <port> ‐‐start
In addition to hand-starting the service, we need to update the BitBake configuration
(for example, build/conf/local.conf) file of each build system, which connects to
server using the PRSERV_HOST variable as described earlier so that each system points
to the server IP and port.
Using package feeds
In order to use package feeds, the following two components have to be
configured properly:
• The server, to provide access to the packages
• The client, to access the server and download the needed packages
The set of packages offered by the package feed is determined by the recipes we
build. We can build one or more recipes and offer them, or we can build a set of
images to generate the desired packages. Once we are satisfied with the amount
of packages offered, we must create the index of packages to be provided by the
package feeds. It is performed by the following command line:
$: bitbake package-index
The package index, along with the packages, must be made available through a
transfer protocol such as HTTP. We can use any server we wish for this task, such as
Apache, Nginx, Lighttpd, and others.
We need to configure the build system to make the runtime package manager, which
is installed in the image, to point to the package feed server. This is done using the
variable PACKAGE_FEED_URIS added in the BitBake configuration (that is, build/
conf/local.conf), similar to the following example:
PACKAGE_FEED_URIS = "http://myfeedserver/repo/"
[ 56 ]
Chapter 6
In addition, our image should have the package-management feature in IMAGE_
FEATURES to ensure that when the image is created for our target, it includes the
package databases and the target-specific tools for runtime package management
to be available on the target. We detail the variable IMAGE_FEATURES in Chapter 11,
Creating Custom Layers.
In case we want a small image with no package management
support, we should not include package-management in
IMAGE_FEATURES.
The PACKAGE_FEED_URIS and IMAGE_FEATURES configurations guarantee that the
image running in the client side can access the server and has the needed utilities in
order to install, remove, and upgrade its packages.
After these steps have been taken, we are able to use the runtime package
management in target. For example, if we choose the RPM package format for the
image, we can fetch the repository information using the following command:
$: smart update
Use the smart query and smart install commands to find and install packages
from the repositories.
The use of package feeds to upgrade the product is a great solution. However, it
must be reliable. Our development process must be very careful and test all different
possible upgrade scenarios in order to avoid bad surprises.
Summary
This chapter presented to us the basic concepts of packaging, a concept that has a
central role for Poky and BitBake, package versioning, and how this impacts
Poky's behavior when rebuilding packages and package feeds. It also showed us
how to configure an image to be updated using prebuilt packages provided by a
remote server.
In the next chapter, we will learn about the BitBake metadata syntax and its
operators to append, prepend, and remove content from variables, variable
expansions, and so on. We will then be able to understand the language used in the
Yocto Project engines better.
[ 57 ]
www.allitebooks.com
Diving into BitBake Metadata
At this point, we know how to generate images and packages and use package feeds,
which is basically everything we must know for simple usage of Poky. Hereafter,
we will learn how to control the behavior of Poky to accomplish our goals and take
maximum benefit from the Yocto Project as a whole.
In this chapter, we will enhance our understanding of the BitBake metadata syntax.
We will learn to use the append, prepend, and remove operators to alter content
from variables, variable expansions, and so on. These are the key concepts we can
use to make our own recipes and customizations that we will learn about in Chapter
10, Exploring External Layers, Chapter 11, Creating Custom Layers, and Chapter 12,
Customizing Existing Recipes.
Using metadata
The amount of metadata used by BitBake is enormous. To take the maximum profit
out of using Poky, we must master it. As we learned in Chapter 4, Grasping the BitBake
Tool, metadata can be classified into the following three major areas:
• Configuration (the .conf files): Configuration files define the global content
that is used to provide information and configure how the classes and recipes
will work
• Classes (the .bbclass files): Classes are available to the whole system and
can be inherited by recipes to easily maintain and avoid code duplication
while the recipes describe the tasks to be run and provide the needed
information to allow BitBake to generate the needed task chain
• Recipes (the .bb or .bbappend files): Recipes and classes are written in a mix
of Python and Shell Scripting code
Diving into BitBake Metadata
Working with metadata
The syntax used by BitBake metadata may be misleading, which can sometimes
be hard to trace. We can check the value of each variable we want using the
environment option (-e or --environment) of BitBake, for example:
$: bitbake -e <recipe> | grep <variable>
In order to understand how BitBake works in more detail, please refer to the
BitBake User Manual.
In the following sections, we will see most of the syntaxes commonly used in recipes.
The basic variable setting
The assignment of a variable can be done as shown:
FOO = "bar"
In the preceding example, the value of the variable FOO is bar.
The variable assignment is core to the BitBake metadata syntax as most other
examples are performed using variables.
Variable expansion
BitBake supports variable referencing. The syntax closely resembles Shell Script,
for example:
A = "aval"
B = "pre${A}post"
This results in A containing aval and B containing preavalpost. One important
thing to bear in mind is that the variable only expands when it is actually used,
as shown:
A = "aval"
B = "pre${A}post"
A = "change"
The preceding example illustrates the laziness of BitBake evaluation. When B is
used, it evaluates A (as it is a reference for it), A now contains change, and B is
prechangepost.
[ 60 ]
Chapter 7
Setting a default value using ?=
When there is a need to provide a default value, the operator ?= can be used. The
following code shows its use:
A ?= "aval"
If A is set before the preceding code is called, it retains its previous value; if A has not
been previously set, it is set to aval. Basically, the ?= operator assigns a new value to
a variable if one has not already been set.
Note that, if there are multiple ?= assignments to a single variable, the first of these is
used. For example, in case we have the following:
A ?= "aval"
A ?= "change"
The variable A has the value aval. However, if we have A previously set, then it is
used. For example:
A = "before"
A ?= "aval"
A ?= "change"
The A variable is kept as before.
Setting a default value using ??=
Another way to provide a default value is using the ??= operator. The difference
between ??= and ?= is that with ??= the assignment does not occur until the end of
the parsing process, so that the last rather than the first ??= assignment to a given
variable is used. Check the following code:
A ??= "somevalue"
A ??= "someothervalue"
If A is set before the preceding code, it retains the value. If A has not been previously
set, it is set to someothervalue.
Immediate variable expansion
Use the := operator when there is a need to force immediate expansion of a variable.
It results in the variable's contents being expanded immediately rather than when
the variable is actually used, as follows:
T = "123"
A := "${B} ${A} test ${T}"
[ 61 ]
Diving into BitBake Metadata
B
T
C
C
= "${T} bval"
= "456"
= "cval"
:= "${C}append"
At the end of this example, A will contain test 123, B will contain 456 bval, and C
will be cvalappend. When A is expanded, B is not yet defined, so B is empty.
Appending and prepending
BitBake offers two sets of appending and prepending operators. The first are += and
=+. The following code illustrates their use:
B = "bval"
B += "additionaldata"
C = "cval"C =+ "test"
At the end of this example, B contains bval additionaldata and C contains test
cval. It is important to note that these operators include an extra space between each
call. When we wish to avoid adding spaces, we use the .= and =. operators. Look at
the following code:
B
B
C
C
= "bval"
.= "additionaldata"
= "cval"
=. "test"
In this example, B is now bvaladditionaldata and C is testcval. In contrast to the
preceding example, the .= and =. operators add no additional space. Commonly, the
+= and =+ operators are used to add items to lists while the .= and =. operators are
used to concatenate strings.
To concatenate strings, we can also use the append and prepend operators, as shown:
B = "bval"
B_append "additionaldata"
C = "cval"
C_prepend "test"
In this example, B is now bvaladditionaldata and C is testcval. These operators
add no additional space.
[ 62 ]
Chapter 7
Conditional metadata set
BitBake provides a very easy-to-use way to write conditional metadata. It is done by
a mechanism called overrides.
The variable OVERRIDES contains values separated by colons (:), and each value is
an item we want to satisfy conditions. So, if we have a variable that is conditional on
arm, and arm is in OVERRIDES, then the version of the variable that is specific to arm is
used rather than the nonconditional version, as shown:
OVERRIDES = "architecture:os:machine"
TEST = "defaultvalue"
TEST_os = "osspecificvalue"
TEST_condnotinoverrides = "othercondvalue"
In this example, TEST will be osspecificvalue due to the condition of os being
in OVERRIDES.
Conditional appending
BitBake also supports appending and prepending to variables based on whether
something is in OVERRIDES, as shown:
DEPENDS = "glibc ncurses"
OVERRIDES = "machine:local"
DEPENDS_append_machine = " libmad"
In the preceding example, DEPENDS is set to glibc ncurses libmad.
File inclusion
BitBake provides two directives for file inclusion: include and require.
With the first directive, include, BitBake attempts to insert the file at that location.
If the path specified on the include line is a relative path, BitBake locates the first
instance it can find within BBPATH. If the referenced file cannot be found, the parsing
does not fail.
By contrast, the second directive, require, raises ParseError if the file to be included
cannot be found. In all other respects, it behaves just like the include directive.
The convention normally adopted in the Yocto Project is to use a .inc
file to share common code between two or more recipe files.
[ 63 ]
Diving into BitBake Metadata
Python variable expansion
BitBake makes it easy to use Python code in variable expansion with the
following syntax:
VARIABLE = "${@<python-command>}"
This gives huge flexibility to the user, as can be seen in the following example:
DATE = "${@time.strftime('%Y%m%d',time.gmtime())}"
This results in the DATE variable containing today's date.
Defining executable metadata
Metadata recipes (.bb) and class (.bbclass) files can use Shell Script code as follows:
do_mytask () {
echo "Hello, world!"
}
This is essentially identical to setting a variable, except that this variable happens to
be an executable shell code. It is important to be careful when writing Shell Script
code as we should not use shell-specific code such as bash or zsh. When in doubt, a
good way to test if our code is safe is to use the dash shell to test it.
Another way to inject code is by using Python code, as shown:
python do_printdate () {
import time
print time.strftime('%Y%m%d', time.gmtime())
}
This is similar to the previous code but flags it as Python so that BitBake knows it is
Python code and runs it accordingly.
Defining Python functions in the global
namespace
We may be required to use Python functions to generate the value for a variable or
some other use. This can be easily done, in recipes (.bb) and classes (.bbclass),
using the following code:
def get_depends(d):
if d.getVar('SOMECONDITION', True):
[ 64 ]
Chapter 7
return "dependencywithcond"
else:
return "dependency"
SOMECONDITION = "1"
DEPENDS = "${@get_depends(d)}"
Usually, we need to access the BitBake database when writing a Python function. A
convention among all metadata is that the d variable is used to point to the BitBake's
database, and it is usually passed as the last parameter of a function.
So, in the preceding example, we ask the database the value of the SOMECONDITION
variable and return a value depending on it.
The example results in DEPENDS containing dependencywithcond.
The inheritance system
The inherit directive is a means of specifying what classes of functionality our recipe
(.bb) requires. It is a rudimentary form of inheritance. For example, we can easily
abstract out the tasks involved in using autoconf and automake, and put that into
.bbclass for our recipes to make use of it. A given .bbclass is located by searching
for classes/filename.bbclass in BBPATH, where filename is what we inherited. So,
in a recipe that uses autoconf or automake, we can use the following:
inherit autotools
This instructs BitBake to inherit autotools.bbclass, so it provides the default tasks
that works fine for most autoconf or automake based projects.
Summary
In this chapter, we learned in detail about the BitBake metadata syntax; its operators
to append, prepend, and remove content from variables; variable expansions; and so
on, including some usage examples for them.
In the next chapter, we will learn how to use Poky to create external compilation
tools and produce a root filesystem suitable for cross development. In addition, the
possible use of Eclipse integration will be explained.
[ 65 ]
Developing with the
Yocto Project
So far, we have used Poky as a build tool; in other words, we have used it as a tool to
design and generate the image that will be used on products. In this chapter, we will
understand how the tool can help us with application or kernel development, creating
external compilation tools, producing a root filesystem suitable for cross development,
and generating an image with the tools for use in the target machine for development.
Deciphering the software development kit
A software development kit (SDK) is a set of tools and files used to develop
and debug. These tools include compilers, linkers, debuggers, external library
headers, and binaries, and may include custom utilities and applications. This set of
programming tools is called a toolchain.
In embedded development, the toolchain is often composed of cross-tools, or tools
executed on one architecture that produces a binary for use in another architecture.
For example, a gcc binary that runs on an x86-64-compatible machine and produces
one binary for an ARM machine is a cross-compiler. When the tool and resultant
binary are executed in the same architecture, it is called a native build.
Usually, when we work on a custom source code and use external libraries, for
example, libusb or libgl, these libraries are used to build at runtime. The custom
source may be built against the library header files, and the binary may be located
somewhere during the execution. The set of files used during build time is placed
under the sysroots directory, part of the Poky SDK, which is very configurable,
depending on application, or it is a very simple one for general use.
Developing with the Yocto Project
When Poky executes tasks through the use of BitBake, it also needs a toolchain to be
able to compile and link binaries for the target. This is called an internal toolchain
because the tools are used internally by the build system. These tools are not
intended to be used externally as they are not prepared for that, although they may
be used on some very specific and advanced use cases.
Working with the Poky SDK
Usually, the standard Poky workflow includes creating an image and package
recipes and deciding what will be installed on the final product image. However, a
huge amount of time is spent developing, writing, testing, or adapting source code
for our application.
So, when we write and test our application, we only care about the application
itself, providing the libraries that the application requires precede application
development, although this can be an iterative process. However, we want a test
environment that looks as similar as possible to the final one, mainly because of the
toolchain compatibility and also to avoid behavioral changes when we integrate the
application into our product.
To help with this task, we can create a toolchain to be used externally with the Poky
environment. Poky generates an SDK package that can be installed on any computer,
independent of Poky being installed on it. In addition, the installed toolchain is
compatible with the internal one. Besides the toolchain, the SDK can also provide a
set of library files (such as library binaries and header files), depending on our needs.
Using an image-based SDK
For a custom source code, we know the dependency libraries and the other
applications we depend on. In such cases, we can create an image that reflects exactly
our needs or uses the closest image provided by Poky.
In order to create the image-based SDK, execute the following command:
$: bitbake core-image-full-cmdline -c populate_sdk
With this command, the SDK is created based on the core-image-full-cmdline
image. If we have a custom image, we can use it instead. The SDK is generated
following the architecture of MACHINE that we have set for.
After it being built, a binary script can be found at build/tmp/deploy/sdk/pokyeglibc-x86_64-core-image-full-cmdline-armv5te-toolchain-1.6.sh.
[ 68 ]
Chapter 8
Do not open the preceding script on a simple text editor.
The script has a piece of binary code that may cause a text
editor to crash.
The resultant script should be installed before being used. We can see the installation
process in the following screenshot:
In the preceding example, the installation directory was /opt/poky/1.6; however,
we may choose any directory. The installation provides the following:
• environment-setup-armv5te-poky-linux-gnueabi: This is the script used
to set up all environment variables needed to use the toolchain
• site-config-armv5te-poky-linux-gnueabi: This is the file with variables
used during the toolchain creation
• version-armv5te-poky-linux-gnueabi: This is the version and
timestamp information
• sysroots: This is a copy of the rootfs directory of images used for the
SDK generation. It includes binary, header, and library files
• armv5te-poky-linux-gnueabi: This contains files for ARM machines
• x86_64-pokysdk-linux: These are files for machines with x86-64
compatibility
Generic SDK – meta-toolchain
Another option is to create a generic SDK, but with cross-compiler, debug tools, and
a basic set of libraries and header files. This generic SDK is called meta-toolchain,
and it is used mainly for kernel and bootloader development and the debug process.
In order to create it, use the following command:
$: bitbake meta-toolchain
[ 69 ]
Developing with the Yocto Project
The resultant file is build/tmp/deploy/sdk/poky-eglibc-x86_64-metatoolchain-armv5te-toolchain-1.6.sh for qemuarm machine. The installation
process is exactly the same as the image-based SDK.
Although this SDK is very helpful, it is highly recommended to create a custom
image that fits our application needs, and then create the SDK based on this.
Using a SDK
To use a SDK to build a custom application, for example, hello-world.c, we can
use the following lines to build it, targeting the ARM architecture:
$: source /opt/poky/1.6/environment-setup-armv5te-poky-linuxgnueabi
$: ${CC} hello-world.c -o hello-world
In order to confirm that the generated binary was properly made for the target
architecture, we can use the file utility as follows:
$: file hello-world
hello-world: ELF 32-bit LSB executable, ARM, EABI5 version 1
(SYSV), dynamically linked (uses shared libs), for GNU/Linux
2.6.16, BuildID[sha1]=1c92a9bb78448a848e88bafa8addf68d6f25460f,
not stripped
Another very commonly used project is the Linux kernel. The Linux kernel uses
the LD utility for linking, so it is necessary to unset the LDFLAGS variable as it is, by
default, defined for use with GCC. When we want to build the Linux kernel source
code, we can use the following sequence of commands:
$: source /opt/poky/1.6/environment-setup-armv5te-poky-linux-gnueabi
$: unset LDFLAGS
$: make defconfig
$: make uImage
Developing applications on the target
For embedded systems, there is a debug portion that should be executed on real
hardware because the application uses hardware-specific peripherals, and it may
be difficult to emulate. In addition, some of the debug processes are dependent on
source signals such as electric or optical, or the resultant action will be a mechanical
behavior that may be difficult to test effectively on emulators.
[ 70 ]
Chapter 8
The first proposed scenario for the development in target is to create a development
image to be used along with an external toolchain. A development image is filled
with header files and additional library links. This will be an image prepared to
provide a build environment for a custom application, and it may be used with a
custom toolchain or the Yocto Project external toolchain. The following line adds
these properties to an image:
IMAGE_FEATURES += "dev-pkgs"
In case we want to modify only build/conf/local.conf, the variable to be used is
EXTRA_IMAGE_FEATURES. The IMAGE_FEATURES variable is better described in Chapter
11, Creating Custom Layers.
The resulting image includes header files and additional library links, and it may be
used during the custom application development cycle. The custom application may
be built against this image root filesystem, which means that the image itself does
not need to be recreated each time. In addition, the image may be shared among all
developers working on the same project. Each one will have a copy, and everyone
will be on the same page.
The dev-pkgs features install all ${PN}-dev packages
into the image.
For developers who prefer, or need, to use native build instead of creating a
development build, Poky can be configured to generate an SDK image. This image
contains the toolchain and development packages (header files and library links).
It means we can have the source code of our custom application being built, run,
tested, and debugged in the target machine.
In order to add the development tools inside an image, we need to include the
tools-sdk feature in the IMAGE_FEATURES variable. We should use EXTRA_IMAGE_
FEATURES if it is added in build/conf/local.conf.
The native build is preferred on newer microprocessors with more processing power
and memory. It may be very slow on older microprocessors.
[ 71 ]
Developing with the Yocto Project
Integrating with Eclipse
Eclipse is a very powerful IDE and widely used for development and debugging
on custom applications. It can be configured to work with Poky SDK. In the Yocto
Project Development Manual, we find the supported Eclipse version and we learn how
to configure it. Included with the manual are the Yocto Project ADT and an image
based on generic toolchain integration.
As soon as our Eclipse is configured, we can use it for development. We can use the
IDE to write the source code, and the Poky toolchain can be used to cross-compile it,
as Eclipse supports the use of this external toolchain.
In addition, we can use Eclipse to deploy the generated binary file to the target,
connected with Eclipse by Ethernet. The binary file and any other needed artifacts are
copied to the target root filesystem, and it is possible to use it right after the transfer.
As soon as the binary is copied to the root filesystem, we can use Eclipse to debug
the application. It means that we are allowed to use step-by-step debug, with the
binary being run directly in the target machine.
In order to accomplish Eclipse integration, it's important to include the Eclipse
feature in the target image by adding the following piece of code in build/conf/
local.conf:
IMAGE_FEATURES += "eclipse-debug"
We, the authors, decided to not include a step-by-step on how to get
Eclipse installed and configured because it requires several steps and
will be outdated very fast. The Yocto Project Development Manual, on
the other hand, gives a canonical place to find a complete and up-todate tutorial for that.
Summary
In this chapter, we learned that the Yocto Project can be used for development as well
as for image creation. We learned how to create two types of toolchains, image-based
and generic, how to use them, and how to create a development image in order to
build and deploy our application in the target. In addition, we learned how we can use
Eclipse in the development phase to write, build, and debug our applications.
In the next chapter, we will understand how we can configure Poky to help us on the
debugging process, how we can configure our system to provide the needed tools for
a remote debug using GDB, how we can track our changes using buildhistory, and
how we can use a handy tool called devshell.
[ 72 ]
Debugging with the
Yocto Project
The debug process is an important step in every development cycle. In this chapter,
we will understand how to configure Poky to help us with the debugging process,
for example, how we can configure our system to provide the needed tools for a
remote debug using GDB, how we can track our changes using buildhistory, and
how we can use a handy tool called devshell.
Differentiating metadata and application
debugging
When we first think about debugging, we usually don't realize that there are
different types of debugging.
Metadata debugging is needed to ensure that the behavior of BitBake's tasks is
aligned with our goals, and to identify the culprit, when it's not. In this case, we
use several log files generated by BitBake in the host in order to help to trace the
execution path of the involved task. As a consequence of a wrong behavior, a file
may not be copied or a feature may not be enabled.
On the other hand, the debugging of run-time code is more natural for us as it is
essentially the same as what we do during the normal development cycle of an
application, a library, or a kernel. Depending on the kind of issue we are after, the
right tool to help may vary from a debugger to code instrumentation (for example,
adding debug prints).
In this chapter, we detail metadata debugging as it is the essence of the Yocto Project
and supports us along the development and use of Poky.
Debugging with the Yocto Project
Tracking image, package, and
SDK contents
The easiest way to ensure we have the image, packages, and SDK, along with the
expected contents, is to use the Build History mechanism. Its contents may change in
unexpected ways when we change a recipe.
When a recipe is updated for a new version or has its code changed, it may influence
the contents put in the generated packages and, as consequence, in the image or SDK.
As Poky deals with a huge amount of recipes, and our images or SDKs frequently have
tens or hundreds of packages included, it may be quite difficult to track the package
contents. The Poky tool responsible to help us in this task is the Build History.
The Build History, as we can guess from its name, keeps the history of contents
of several artifacts built during the Poky use. It can track package, image, and
SDK building.
To enable the Build History in our system, we need to add the following lines of code
in our local.conf file:
INHERIT += "buildhistory"
BUILDHISTORY_COMMIT = "1"
The INHERIT method includes the buildhistory class hooks into the building
process, while the BUILDHISTORY_COMMIT line enables BitBake to create a new Git
commit in the buildhistory repository for every new package, image, or an SDK
build. This makes the track as simple as git diff between two commits.
The data for all packages, images, and SDKs built is stored under the build/
buildhistory directory as text files so that it is easy to use this data to extract
extra information. Poky provides a utility that outputs the difference between two
buildhistory states, called buildhistory-diff, in a more concise way, which is
very useful when checking for changes.
The buildhistory-diff utility outputs the difference between any two Git
revisions in a more meaningful way. We can see an example of its output in the
following screenshot:
[ 74 ]
Chapter 9
The preceding screenshot shows the differences highlighted by buildhistory-diff
when the package strace is added in the core-image-minimal image.
When a package is built, buildhistory creates a list of generated subpackages,
installation scripts, a list of file ownership and sizes, the dependency relation, and
more. For images and SDKs, the dependency relationship among the packages,
filesystem files, and graph of dependency is created.
For a better understanding of all capabilities and features provided by the build
history, you are advised to read the Yocto Project Reference Manual.
Debugging packaging
In more sophisticated recipes, we split the installed contents in several subpackages.
The subpackages can be optional features, modules, or any other set of files that are
optional to be installed.
To inspect how the recipe's content has been split, we can use the build/tmp/
work/<arch>/<recipe name>/<software version>/packages-split directory.
It contains a subdirectory for every subpackage and has its contents in the subtree.
Among the possible reasons for a mistaken content split, we have the following:
• Contents not being installed (for example, an error in installation scripts)
• Application or library configuration error (for example, a disabled feature)
• Metadata error (for example, wrong package order)
Another common issue that we find, mainly in library recipes, is that the needed
artifacts are not made available in the sysroot directory (for example, headers
or dynamic libraries), causing a build breakage. The counterpart of the sysroot
generation can be seen at build/tmp/work/<arch>/<recipe name>/<software
version>/sysroot-destdir.
[ 75 ]
Debugging with the Yocto Project
Other possible causes may need us to instrument the task code with the further
presented logging functions, so we can figure out the logical error or bug that causes
the unexpected result.
Logging information during task
execution
The logging utilities provided by BitBake are very useful to trace the code execution
path. BitBake provides logging functions for use in Python and Shell Script code,
as described:
• Python: For use within Python functions, BitBake supports several log levels,
which are bb.fatal, bb.error, bb.warn, bb.note, bb.plain, and bb.debug
• Shell Script: For use in Shell Script functions, the same set of log levels exists
and is accessed with a similar syntax: bbfatal, bberror, bbwarn, bbnote,
bbplain, and bbdebug
These logging functions are very similar to each other but have inner differences
as described:
• bb.fatal and bbfatal: These have the highest priority of logging messages
as they print the message and terminate the processing. They cause the build
to be interrupted.
• bb.error and bberror: They are used to display an error but do not force
the build to stop.
• bb.warn and bbwarn: They are used to warn users about something.
• bb.note and bbnote: These add a note to the user. They are only informative.
• bb.plain and bbplain: These output a message.
• bb.debug and bbdebug: These add debugging information that is shown
depending on the debug level used.
There is one subtle difference between the use of the logging functions in Python and
Shell Scripting.
The logging functions in Python are directly handled by BitBake, seen on the
console, and stored in the execution log that can be seen inside build/tmp/log/
cooker/<machine>.
When the logging functions are used in Shell Script, the information is
outputted on the task's respective task log file, which is available in build/tmp/
work/<arch>/<recipe name>/<software version>/temp.
[ 76 ]
Chapter 9
Inside it, we can run the scripts for every task with the pattern run.<task>.<pid> and
use the pattern log.<task>.<pid> for its output. For convenience, symbolic links are
kept updated by BitBake, pointing to the last log files using the pattern log.<task>;
thus, we can actually check for log.do_compile, for example, when intending to
verify whether the right files were used during the build process. Alternatively, we can
check for log.do_patch to verify whether a patch has been applied.
The build/tmp/work directory is detailed in Chapter 5, Detailing the Temporary
Build Directory.
Utilizing a development shell
When editing packages or debugging build failures, a development shell can be
a useful tool. When we use devshell, source files are extracted into the working
directory, patches are applied, a new terminal is opened, and files are placed in the
working directory.
In the new terminal, all the environment variables needed for the build are still
defined, so we can use commands such as configure and make. The commands
execute just as if the build system was executing them.
The following command is an example that uses devshell on a target named
linux-yocto:
$: bitbake linux-yocto -c devshell
This allow us to rework the Linux kernel source code and build it in place, to avoid
building it from scratch in our development machine, and change its code as needed.
It is important to bear in mind that all changes made inside devshell
are not persistent between builds; thus, we must be careful to record
any change that is important, prior to leaving it.
As we have the source at our disposal, we can use it to generate extra patches. One
very practical way of doing that is using Git and git format-patch to create the
patch to be included in the recipe afterwards.
In order to include the generated patch in the recipe and make it persistent, see
Chapter 12, Customizing Existing Recipes.
[ 77 ]
Debugging with the Yocto Project
Using the GNU Project Debugger for
debugging
While developing any project we, from time to time, end up struggling to
understand subtle bugs. The GNU Project Debugger (GDB) is available as a
package within Poky and is installed in SDK images by default, as was detailed in
Chapter 8, Developing with the Yocto Project.
In order to install debugging packages that contain the debug symbols
and debugging tools in an image, add IMAGE_FEATURES += "dbgpkgs tools-debug" in local.conf.
The use of the SDK, or an image with the debugging packages and tools installed,
allows us to debug applications directly in the target, as we usually do in our
development machine.
Sometimes, due to memory or disk space constraints, it is not possible to use GDB
directly on the target to debug. These constraints arise because GDB needs to load
the debugging information and the binaries of the process being debugged, and
it needs to perform many computations to locate information such as function
names, variable names and values, stack traces, and so forth even before starting the
debugging process.
To overcome these constraints, we can use gdbserver, included by default when
using tools-debug in IMAGE_FEATURES. It runs on the target and does not load
any debugging information from the debugged process. Instead, a GDB instance
processes the debugging information on the host. The host GDB then sends control
commands to gdbserver to control the debugged application.
As the host GDB is responsible for loading the debugging information and
performing the necessary processing to make the debugging process take place,
the target does not need to have the debugging symbols installed and we need to
make sure the host can access the binaries with their debugging information. It is
advisable that the target binaries are compiled with no optimizations to facilitate the
debugging process.
The process to use gdbserver and configure the host and target in the appropriate
way is detailed in the Yocto Project Development Manual.
[ 78 ]
Chapter 9
Summary
In this chapter, we understood how we can configure Poky to help us with the
debugging process. We learned the contents of deployed directories that can be used
on debugging, how we can track our changes using buildhistory, how we can use
devshell to emulate the same build environment found by BitBake, and how we
configure our system to provide the needed tools for GDB debugging.
In the next chapter, we will understand how to expand the Poky source code using
external layers. We will be introduced to the concept of layering, and we will learn in
detail the directory structure and the content of each layer type; additionally, we
will learn how to add an external layer on our project manually or by using the
Hob configuration.
[ 79 ]
Exploring External Layers
One of the most charming features of Poky is the flexibility of using external layers.
In this chapter, we understand why this is a strong capability and how we can
take advantage of this. We understand layer types and how their directory trees
work. Finally, we learn how to include a new layer on our project or on a
Hob configuration.
Powering flexibility with layers
Poky contains a great amount of metadata as machine definition files, classes, and
recipes that cover everything from simple applications to full graphical stacks
and frameworks. The main motivation to use layers is to organize the long list of
providers better and still make sure users may be able to pick only the needed or
desired provider. It is also a way to provide a customizable source code that can be
used for any architecture or modified in the way the user needs.
We can choose the set of layers needed for each project. We can configure the provided
source code for our needs because we want to configure how one package acts in our
product or because we want the package built with our architecture flags. This means
that using layers, we can provide additional features to Poky (represented by recipes
from additional layers) that may be modified according to our needs (represented by
the configuration appends we include for external recipes).
In addition, when creating our own custom project environment, instead of changing
recipes and configuration files and appending files into one single layer, we ought
to organize the metadata in different layers. The more separated the organization is,
the easier it is to reuse the layers in future for other projects; for example, the Poky
source code is separated in different layers. By default, it has three included layers as
we can see on the output of the following command line:
$: bitbake-layers show-layers
Exploring External Layers
The result can be seen in the following screenshot:
The command-line output shows the following three important properties of
any layer:
• Name: This usually starts with the string meta.
• Path: This is important when we want to add an additional layer in our
project being appended to the variable BBPATH.
• Priority: This is the value used by BitBake to decide which recipe to use and
the order in which the .bbappend files should be joined. It means that if two
layers include the same recipe file (.bb), the one with higher priority is used.
In the .bbappend case, every .bbappend file is included in the original recipe,
and the layer priority determines the order of inclusion so the .bbappend files
within the highest priority layers are appended first, followed by the others.
Poky is arranged in three individual layers, coincidentally the three types available.
The layer meta is the OpenEmbedded Core metadata, which contains the recipes,
classes, and the QEMU machine configuration files. It is a software layer.
A software layer includes only applications or configuration files for applications
and can be used on any architecture. There is a huge list of software layers. To name
only a few, we have meta-java, meta-qt5, and meta-browser. The layer meta-java
provides Java runtime and SDK support, the layer meta-qt5 includes Qt5 support,
and meta-browser supports several web browsers such as Firefox and Chrome.
The layer meta-yocto-bsp is the Poky reference Board Support Package (BSP)
layer. It contains machine configuration files and recipes to configure packages for
the machines. As it is a reference BSP layer, it can be used as an example.
[ 82 ]
Chapter 10
The layer meta-yocto is the Poky reference distribution layer. It contains
a distribution configuration used in Poky by default; this is an example of a
distribution file. This default distro is described in the poky.conf file, and it is
widely used for testing products. However, sometimes your product may have
special needs and changes in build/conf/local.conf as required. The build/
conf/local.conf file is a volatile file that is not supposed to be tracked by Git. We
should not rely on it to set package versions, providers, and the system features for
products, but use it just as a shortcut to test purposes during development.
The most adequate and maintainable solution is to create a distribution and a
distribution layer to place the distro file into. This configuration allows any build to
be reproduced afterwards. We can use one distribution layer for all distributions we
have, or we can create one new layer for every new distribution; the better approach
depends on your project needs.
The policy configuration provided in a distribution layer overrides
the same configuration from build/conf/local.conf.
Detailing the layer's source code
Usually, a layer has a directory tree, as shown in the following figure:
meta-* /
classes
conf
COPYING
README
recipes-*
The layer name should start with meta-, it is not a requirement but the advised
pattern. Inside this directory, there are two files, <layer>/COPYING and <layer>/
README, a license, and a message to the user. In <layer>/README, we must specify
any other dependency and information that the layer's users need to know.
The classes folder should hold the classes provided and specific to that layer (the
.bbclass files). It is an optional directory.
[ 83 ]
Exploring External Layers
The folder <layer>/conf is mandatory and should provide the configuration files
(the .conf files). Primarily, the layer configuration file <layer>/conf/layer.conf,
to be detailed in the next chapter, is the file with the layer definition.
When the <layer>/conf folder is from a BSP layer, the directory structure should
look like the following figure:
conf /
layer .conf
machine
include
<another>.conf
<machine-name>.conf
<other>.conf
If the <layer>/conf folder is from a distribution layer, the directory structure
should look like the following figure:
conf /
distro
<distro-name>.conf
include
<another>.conf
layer . conf
<other>.conf
The recipe-* folder is a cluster of recipes separated by category, for example,
recipes-core, recipes-bsp, recipes-graphic, recipes-multimedia, and
recipes-kernel. Inside each folder starting with recipes-, there is a directory
with the recipe name or a group of recipes; inside it, the recipe files end with .bb or
.bbappend. For example, we can find the following resumed figure from meta:
[ 84 ]
Chapter 10
recipes-multimedia /
alsa
alsa-fpu. inc
alsa-lib
alsa-lib_1.0.27.2.bb
alsa-tools
alsa-tools_1.0.26.1.bb
alsa-utils
alsa-utils_1.0.27.2.bb
alsa-utils-alsaconf_1.0.27.2.bb
flac
flac-1.3.0
Adding meta layers
There are hundreds of meta layers from the Yocto Project, OpenEmbedded,
communities, and companies that should be manually cloned inside the project
source directory to be used. We can find them at http://git.yoctoproject.org/
or http://layers.openembedded.org.
In order to include, for example, meta-realtime in our project, we can change the
content of configuration for Layers; click on Add layer, browse to the meta-realtime
folder, and click on OK if we use Hob, as shown in the following screenshot:
We notice that the layers meta, meta-yocto, and meta-hob cannot be removed as
they provide the core infrastructure and BSP for Hob. These layers are the Yocto
Project definition, plus the Hob provider layer.
[ 85 ]
Exploring External Layers
If we use the BitBake command lines, we can add a new layer by changing the file
conf/bblayer.conf and add the absolute path to the new meta layer directory, as
shown in the following source code. The line highlighted is the one to be added. The
others are the default values for this file.
In the next BitBake command, or Hob launch, the added layer is parsed and the
meta-realtime metadata is included in BitBake's database, allowing the packages
inside the added layer to be used.
Summary
In this chapter, we introduced the concept of layering. We learned in detail the
directory structure and the content of each layer type. In addition, we saw how to
add an external layer on our project manually or by using the Hob configuration.
In the next chapter, we will learn more about why we need to create new
layers and what is the common metadata included in them (such as machine
definition files, recipes, and images), and wrap it all up with an example of
distribution customization.
[ 86 ]
Creating Custom Layers
Beyond using existing layers from the community or from vendors, we learn why
we create layers for our own products in this chapter. In addition, we understand
how to create a machine definition and a distribution, and profit from them to better
organize our source code.
Making a new layer
Before creating our own layer, it's always a good idea to check if there is a similar
one already available in the OpenEmbedded metadata index (http://layers.
openembedded.org). If we cannot find any suitable layer for our needs, the next step
is to create the directory. Usually, the layer name starts with meta-, but this is not a
technical restriction.
The layer configuration file is required in every layer and is placed in <layer>/
conf/layer.conf; we can create it manually using any text editor or populate it
with a script provided in Poky, as shown in the following command:
$: ./poky/scripts/yocto-layer create newlayer
The output is shown in the following screenshot:
Creating Custom Layers
With the script, we are asked to enter the value for layer priority and answer other
questions regarding the sample content that can be generated for the layer. We can
use the default values or enter a custom one. An example of a generated layer is
shown in the following figure:
meta-newlayer/
conf
layer.conf
COPYING.MIT
README
recipes-example
example
example-0.1
example.patch
helloworld.c
example_0.1.bb
recipes-example-bbappend
example-bbappend
example-0.1
example.patch
example_0.1.bbappend
Important variables that we may need to add in case our layer requires other layers
to work are as follows:
• LAYERVERSION: This is an optional variable that specifies the version of the
layer in a single number. This variable is used within the LAYERDEPENDS
variable in order to depend on a specific version of a layer, and it must be
suffixed with the layer's name, for example, LAYERVERSION_newlayer = "1".
• LAYERDEPENDS: This lists the layers that the recipes depend upon,
separated by spaces. Optionally, we can assign a specific layer version for
a dependency by adding it to the end of the layer name with a colon, for
example, otherlayer:2. This variable must be suffixed with the name of the
specific layer, for example, LAYERDEPENDS_newlayer = "otherlayer".
If a dependency cannot be satisfied or the version numbers do not match, an error is
raised. The base of the layer structure is now created and we see how to extend it in
the following sections.
[ 88 ]
Chapter 11
Adding metadata to the layer
The reasoning behind the use of layers is to add extra metadata to BitBake's database
or change it.
The most commonly added features are project related, such as applications,
libraries, or a service server.
On the other hand, instead of adding new features, it is much more usual to
accommodate existing feature configurations to our needs, for example, the initial
network values for a SSH server or the boot splash picture.
It means that we can include several types of metadata files on a new layer, recipes,
images, and bbappend files to change existing features. The script used to create the
new layer can, as well, create two example files; the first one, example_0.1.bb is a
recipe example. The second one, example_0.1.bbappend, is a bbappend example
used to modify the feature included by example_0.1.bb. There are several other
examples of bbappend files on meta-yocto-bsp or meta-yocto, and we explore
some of their common usages in the next chapter.
Creating an image
The image files can be seen as a set of packages grouped for a purpose and
configured in a controlled way. We can create a new image, including an existing
image, add the needed packages or override configurations; or we can create the
image from scratch.
When an image mostly fits our needs and we need to do minor adjustments on
it, it is very convenient to reuse its code. This makes code maintenance easier
and highlights the functional differences. For example, if we want to include an
application and remove an image feature, we can create an image at recipes-my/
images/my-image-sato.bb with the following lines of code:
require recipes-sato/image/core-image-sato.bb
IMAGE_FEATURES_remove = "splash"
CORE_IMAGE_EXTRA_INSTALL += "myapp"
On the other hand, we sometimes want to create our image from scratch; we can
facilitate our work using the core-image class as it provides a set of image features
that can be used very easily, for example, an image in recipes-my/images/myimage-nano.bb consists of the following lines of code:
inherit core-image
IMAGE_FEATURES += "ssh-server-openssh splash"
CORE_IMAGE_EXTRA_INSTALL += "nano"
[ 89 ]
Creating Custom Layers
The append operator (+=) is used to guarantee that a new IMAGE_
FEATURES variable can be added by build/conf/local.conf.
CORE_IMAGE_EXTRA_INSTALL is the variable we should use to include extra packages
into the image when we inherit the core-image class that facilitates image creation,
adding support for the IMAGE_FEATURES variable, which avoids a lot of duplication
of code. The IMAGE_INSTALL variable groups the CORE_IMAGE_EXTRA_INSTALL
contents, and IMAGE_FEATURES related packages generate our root filesystem.
Currently, the following are the image features supported:
• dbg-pkgs: This installs debug symbol packages for all packages installed in a
given image.
• dev-pkgs: This installs development packages (headers and extra library
links) for all packages installed in a given image.
• doc-pkgs: This installs documentation packages for all packages installed in
a given image.
• nfs-server: This installs an NFS server.
• read-only-rootfs: This creates an image whose root filesystem is
read-only.
• splash: This enables showing a splash screen during boot. By default, this
screen is provided by psplash, which allows customization. If you prefer to
use an alternative splash screen package, you can do so by setting the SPLASH
variable to a different package name (or names) within the image recipe or at
the distro configuration level.
• ssh-server-dropbear: This installs the Dropbear minimal SSH server that
is used by default in case we use ssh-server, as it offers a smaller footprint
and most of the OpenSSH features.
• ssh-server-openssh: This installs the OpenSSH SSH server, which is more
full-featured than Dropbear. Note that if both the OpenSSH SSH server and
the Dropbear minimal SSH server are present in IMAGE_FEATURES, then
OpenSSH will take precedence and Dropbear will not be installed.
• staticdev-pkgs: This installs static development packages (that is, static
libraries containing the *.a files) for all packages installed in a given image.
• tools-debug: This installs debugging tools such as strace and gdb.
• tools-profile: This installs profiling tools such as oprofile, exmap,
and LTTng.
[ 90 ]
Chapter 11
• tools-sdk: This installs a full SDK that runs on the device.
• tools-testapps: This installs device testing tools (for example,
touchscreen debugging).
• x11: This installs the X server.
• x11-base: This installs the X server with a minimal environment.
• x11-sato: This installs the OpenedHand Sato environment.
Adding a package recipe
A package recipe is how we can instruct BitBake to fetch, unpack, compile, and
install our application, kernel module, or any software provided by a project. Poky
includes several classes that abstract the process for the most common development
tools as projects based on Autotools, CMake, and QMake. A list of classes included
in Poky can be seen in the Yocto Project Reference Manual.
One simple recipe that does the compile and install tasks explicitly is provided
as follows:
DESCRIPTION = "Simple helloworld application"
SECTION = "examples"
LICENSE = "MIT"
LIC_FILES_CHKSUM = "file://${COMMON_LICENSE_DIR}/MIT;md5=0835ade698e0b
cf8506ecda2f7b4f302"
SRC_URI = "file://helloworld.c"
S = "${WORKDIR}"
do_compile() {
${CC} helloworld.c -o helloworld
}
do_install() {
install -d ${D}${bindir}
install -m 0755 helloworld ${D}${bindir}
}
The do_compile and do_install code blocks provide the Shell Scripting commands
to build and install the resulting binary into the destination directory that is
referenced as ${D}.
[ 91 ]
Creating Custom Layers
However, in the case of an Autotools-based project, we can avoid a lot of code
duplication using the autotools class in the stripped example extracted from the
recipe poky/meta/recipes-core/dbus-wait/dbus-wait_git.bb, as follows:
DESCRIPTION = "A simple tool to wait for a specific signal over DBus"
...
inherit autotools
The simple act of inheriting the class is in fact providing all the code required to do
the following tasks:
• Updating the configure script code and artifacts
• Updating the libtool scripts
• Running the configure script
• Running make
• Running make install
The same concepts apply to other building tools, as is the case for CMake and QMake.
The number of supported classes is growing, and it is expected that new ones will be
included in every release to support new build systems and avoid code duplication.
Writing a machine definition
Creating a new machine to be used by Poky is a straightforward task. It essentially
provides the information needed for a machine to work. The bootloader, kernel, and
hardware support drivers must be checked before starting to integrate the board into
the BSP layer.
The Yocto Project supports x86-32, x86-64, ARM32, ARM64, MIPS, MIPS64, and
PowerPC, representing the most currently used embedded architectures.
The prevailing set of variables used in a machine definition is as follows:
• TARGET_ARCH: This sets the machine architecture, for example, ARM
and i586
• PREFERRED_PROVIDER_virtual/kernel: This overrides the default kernel
(linux-yocto) in case you need to use a specific one
• SERIAL_CONSOLES: This defines serial consoles and their speeds
• MACHINE_FEATURES: This describes hardware features, so the needed
software stack is included in the images by default
[ 92 ]
Chapter 11
• KERNEL_IMAGETYPE: This is used to choose the kernel image type, for
example, zImage and uImage
• IMAGE_FSTYPES: This sets the generated filesystem image types, for example,
tar.gz, ext4, and ubifs
You can see examples of machine definition files inside the Poky source code in
meta-yocto-bsp/conf/machine/.
When describing a new machine, we should pay special attention to specific features
supported by it in MACHINE_FEATURES. This way, the needed software to support
these features is installed into the images. The current available values for MACHINE_
FEATURES are listed as follows:
• acpi: This indicates that the hardware has ACPI (x86/x86_64 only)
• alsa: This indicates that the hardware has ALSA audio drivers
• apm: This indicates that the hardware uses APM (or APM emulation)
• bluetooth: This indicates that the hardware has integrated Bluetooth
• ext2: This indicates that the hardware has HDD or microdrive
• irda: This indicates that the hardware has an IrDA support
• keyboard: This indicates that the hardware has a keyboard
• pci: This indicates that the hardware has a PCI bus
• pcmcia: This indicates that the hardware has PCMCIA or
CompactFlash sockets
• screen: This indicates that the hardware has a screen
• serial: This indicates that the hardware has serial support (usually RS232)
• touchscreen: This indicates that the hardware has a touchscreen
• usbgadget: This indicates that the hardware is compatible with USB
gadget devices
• usbhost: This indicates that the hardware is compatible with the USB host
• wifi: This indicates that the hardware has integrated Wi-Fi
Using a custom distribution
The creation of a distribution is a mix of simplicity and complexity. The process of
creating the distribution file is very easy; however, the distribution configuration has
a high impact in the way Poky behaves and may cause a binary incompatibility with
previously built binaries, depending on the options we use.
[ 93 ]
Creating Custom Layers
The distribution is where we define global options such as the toolchain version,
graphical backends, support for OpenGL, and so on. We should make a distribution
only in case the default settings provided by Poky do not fulfill our requirements.
Usually, we intend to change a small set of options from Poky. For example, we
remove the X11 support to use framebuffer instead. We can easily accomplish this
reusing Poky distribution and overriding the variables we need. For example, the
sample distribution represented by the file <layer>/conf/distro/mydistro.conf
is as follows:
require conf/distro/poky.conf
DISTRO = "mydistro"
DISTRO_NAME = "mydistro (My New Distro)"
DISTRO_VERSION = "1.0"
DISTRO_CODENAME = "codename"
SDK_VENDOR = "-mydistrosdk"
SDK_VERSION := "${@'${DISTRO_VERSION}'.replace('snapshot${DATE}','snapshot')}"
MAINTAINER = "mydistro <mydistro@mycompany.com>"
DISTRO_FEATURES_remove = "x11"
To use the distribution just created, we need to add the following piece of code in
build/conf/local.conf:
DISTRO = "mydistro"
The variable DISTRO_FEATURES may influence how the recipes are configured and
the packages are installed in images. For example, if we want to be able to use sound
in any machine and image, the alsa features must be present. The following list
shows the present state for DISTRO_FEATURES supported values:
• alsa: This includes ALSA support (OSS compatibility kernel modules
installed, if available).
• bluetooth: This includes Bluetooth support (integrated BT only).
• cramfs: This includes CramFS support.
• ext2: This includes tools to support devices with internal HDD/microdrives
to store files (instead of flash-only devices).
• ipsec: This includes IPSec support.
• ipv6: This includes IPv6 support.
• irda: This includes IrDA support.
[ 94 ]
Chapter 11
• keyboard: This includes keyboard support (for example, keymaps will be
loaded during boot).
• nfs: This includes NFS client support (to mount NFS exports on devices).
• opengl: This includes the Open Graphics Library, a cross-language,
multiplatform application programming interface used for rendering twoand three-dimensional graphics.
• pci: This includes PCI bus support.
• pcmcia: This includes the PCMCIA/CompactFlash support.
• ppp: This includes PPP dialup support.
• smbfs: This includes SMB networks client support (to mount Samba/
Microsoft Windows shares on devices).
• systemd: This includes support for the init manager, a full replacement of
init with parallel starting of services, reduced shell overhead, and other
features. This init manager is used by many distributions.
• usbgadget: This includes USB gadget device support (for USB networking/
serial/storage).
• usbhost: This includes USB host support (allows the connection of an
external keyboard, mouse, storage, network, and so on).
• wayland: This includes the Wayland display server protocol and the library
that supports it.
• wifi: This includes Wi-Fi support (integrated only).
Machine features versus distro features
Both DISTRO_FEATURES and MACHINE_FEATURES work together to provide feasible
support on the final system.
When a machine supports a feature, this does not imply it is being supported by the
final system because the distribution used must provide the underlying base for it.
If a machine supports Wi-Fi but the distribution does not, the applications used by
the operating system will be built with Wi-Fi support disabled so that the outcome
will be a system without Wi-Fi support.
On the other hand, if the distribution provides Wi-Fi support and a machine does
not, the modules and applications needed for the Wi-Fi will not be installed in
images built for this machine, though the operating system and its modules have
support for Wi-Fi enabled.
[ 95 ]
Creating Custom Layers
Understanding the variables scope
The BitBake metadata has thousands of variables, but the scope where these
variables are available depends on where it is defined. Basically, there are two kinds
of variables as follows:
• Variables defined in configuration files are global to every recipe. The
parsing order of the main configuration files is shown as follows:
°°
build/conf/local.conf
°°
<layer>/conf/machines/<machine>.conf
°°
<layer>/conf/distro/<distro>.conf
• Variables defined within recipe files are local to the specific recipe only
during the execution of its tasks.
Summary
In this chapter, we learned the reasons that motivate us to create a new layer
and metadata. We saw a description on how to create machine configuration, a
distribution definition, and recipes files. We learned how we can create images and
how to include our application to an image.
In the next chapter, we will access some examples of the most common
customization cases used by an additional layer, such as modifying existing
packages, adding extra options to autoconf, applying a new patch, and including a
new file to a package. We will see how to configure BusyBox and linux-yocto, two
packages commonly customized when making an embedded system.
[ 96 ]
Customizing Existing Recipes
In the course of our work with Yocto Project's tools, it is expected that we need
to customize existing recipes. In this chapter, we explore some examples, such as
changing compilation options, enabling or disabling features of a recipe, applying an
extra patch, and modifying BusyBox and Linux Yocto Framework settings.
Common use cases
Nowadays, projects usually have a set of layers to provide the required features; we
certainly need to make changes on top of them, to adapt them to our specific needs.
It may be cosmetic or substantive changes, but the way to accomplish them is
the same.
To make changes to a preexisting recipe, we need to create a .bbappend file in our
project layer. The name of the file is the same as the original recipe along with the
append suffix. For example, if the original recipe was named <original-layer>/
recipes-core/app/app_1.0.bb, our respective .bbappend will be <layer>/
recipes-core/app/app_1.0.bbappend.
The .bbappend file can be seen as a piece of text that is appended at the end of the
original recipe. It empowers us with an extremely flexible mechanism to avoid
duplicating source code in order to apply the needed changes in our project's layers.
When there is more than one .bbappend file for a recipe, all of them
are joined following the layer priority order.
Customizing Existing Recipes
Adding extra options to recipes based
on Autoconf
Let's assume we have Autoconf's build system-based application with a preexisting
recipe for it, and we want to do the following:
• Enable my-feature
• Disable another-feature
The content of the .bbappend file, in order to make the changes, will be
the following:
EXTRA_OECONF += "--enable-my-feature --disable-another-feature"
EXTRA_OEMAKE += "DEFINE_PASSED_TO_MAKE=1"
The same can also be done based on the hardware we are building for, as follows:
EXTRA_OECONF_append_arm = " --enable-my-arm-feature"
EXTRA_OEMAKE_append_mymachine = " MYMACHINE_SPECIFIC=1"
EXTRA_OECONF is used to add extra options to the configure script.
EXTRA_OEMAKE is used to add extra parameters to the make call.
Applying a patch
For cases when we need to apply some patch over an existent package, we should
use FILESEXTRAPATHS, which includes new directories into the searching algorithm,
making the extra file visible to BitBake, as shown:
FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}-${PV}:"
SRC_URI += "file://mypatch.patch"
In the preceding example, the current directory / package version
(${THISDIR}/${PN}-${PV}:) is included in the directories list used for file
searching. The use of the _prepend operator is important as it guarantees that our
provided file is used, even if a file with the same name is added in the lower priority
layers in future.
BitBake assumes that every file with a .patch extension is a patch and applies
it accordingly.
[ 98 ]
Chapter 12
Adding extra files to the existing packages
If we need to include an extra configuration file, we should use FILESEXTRAPATHS, as
explained in the previous example and shown in the
following lines of code:
FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}-${PV}:"
SRC_URI += "file://newconfigfile.conf"
do_install_append() {
install -m 644 ${WORKDIR}/newconfig.conf ${D}${sysconfdir}
}
The do_install_append function appends the provided block below the metadata
already available in the original do_install function. It includes the command
needed to copy our new configuration file into the package filesystem. The file
is copied from ${WORKDIR} to ${D} as the directory used by Poky to build the
package and the destination directory used by Poky to create the package. The
${sysconfdir} directory is the system configuration directory (usually with /etc).
We should use the variables provided on top of poky/meta/
conf/bitbake.conf instead of pointing to hardcoded
paths. For example, use ${sysconfdir} instead of /etc and
${bindir} in place of /usr/bin.
Understanding file searching paths
When a file (a patch or a generic file) is included in SRC_URI, BitBake searches it
in the variables FILESPATH and FILESEXTRAPTH. The default setting is to look in
the following:
• <recipe>-<version>/
• <recipe>/
• files/
Besides this, it also checks for OVERRIDES for an override's specific file in each place.
To illustrate this, consider a recipe foo_1.0.bb and the variable OVERRIDES =
"<board>:<arch>"; the file will be searched in the following:
• foo-1.0/<board>/
• foo-1.0/<arch>/
• foo-1.0/
[ 99 ]
Customizing Existing Recipes
• foo/<board>/
• foo/<arch>/
• foo/
• files/<board>/
• files/<arch>/
• files/
This is just illustrative as the list of OVERRIDES is huge and machine-specific. When
we work with our recipe, we can use bitbake -e in order to know the full list of
available overrides for a specific machine and use it accordingly.
Changing recipe feature configuration
One supported mechanism to ease feature-set customization for recipes is
PACKAGECONFIG. It provides a way to enable and disable the recipe features. For
example, if the recipe has the following configuration:
PACKAGECONFIG ?= "feature1"
PACKAGECONFIG[feature1] = "--enable-feature1,--disablefeature1,feature1depends"
PACKAGECONFIG[feature2] = "--enable-feature2,--disablefeature2,feature2depends"
The recipe has two features, feature1 and feature2. For each configuration option,
there is a string to define how to enable the feature on autoconf, how to disable the
feature on autoconf, and the new dependencies in case the option is enabled.
We can create a .bbappend file that expands the PACKAGECONFIG variable's default
value to enable feature2 as well, as shown:
PACKAGECONFIG += "feature2"
In order to add the same feature in the build/conf/local.conf
file, we can use PACKAGECONFIG_pn-<recipename>_append =
'feature2'.
More detailed information about the use of PACKAGECONFIG and its options can be
found in the Yocto Project Reference Manual.
[ 100 ]
Chapter 12
Customizing BusyBox
BusyBox is a key component of most embedded Linux-based projects as it
provides an alternative with a smaller footprint when compared to the usual Linux
counterparts for the most commonly used utilities. It can be seen as a Swiss knife
since it provides a huge set of utilities and is quite flexible regarding which utilities
to enable or disable. Poky provides a default setting for BusyBox, and it sometimes
may not fulfill our needs; so, changing this configuration is a common task. For
example, the <layer>/recipes-core/busybox/busybox_1.22.1.bbappend file
could have the following lines of code:
FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
SRC_URI += "file://enable-tftpd.cfg"
The <layer>/recipes-core/busybox/busybox/enable-tftpd.cfg file contains
the following:
CONFIG_TFTPD=y
This combination of the .bbappend file and configuration file is enough to enable
support for the TFTP server in BusyBox.
For the cases when we want to deselect an option, we can add a
negative line instead, for example, CONFIG_TFTPD=n.
Customizing the linux-yocto framework
The Linux kernel is a complex software that provides an infinite number of
possible configurations. The Yocto Project provides a framework (linux-yocto)
to manage a huge set of machines in a single kernel tree. We can take advantage
of this framework to enable or disable features for our machine, for example, by
using <layer>/recipes-kernel/linux/linux-yocto_3.14.bbappend with the
following content:
FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:"
SRC_URI += "file://enable-can.cfg"
The content of the <layer>/recipes-kernel/linux/linux-yocto/linux-yocto/
enable-can.cfg file is as shown:
CONFIG_CAN=y
[ 101 ]
Customizing Existing Recipes
One common requirement when doing a Linux-based embedded system is to change
the kernel configuration. We can do this using the SDK or BitBake, as explained:
• Using SDK: The creation and installation of the Yocto Project's SDK is
detailed in Chapter 8, Developing with the Yocto Project. After having the SDK
exported, we can configure the Linux kernel source in the usual way (for
example, make menuconfig).
• Using BitBake: When small changes or testing is needed, we can use BitBake
to configure or generate the Linux kernel configuration file. We can use the
following command lines to achieve this:
$: bitbake virtual/kernel -c menuconfig
$: bitbake virtual/kernel -c savedefconfig
For Linux kernel development, the use of SDK is preferred as it provides a convenient
development environment; BitBake should be used only for quick changes.
We need to bear in mind that not every supported machine in the Yocto
Project, vendor, and community BSP layers use the linux-yocto
framework. This means that the configuration fragment mechanism
is not available for those machines, and we need to provide a full
defconfig file in order to customize their kernel.
The complete linux-yocto documentation can be found in the Yocto Project Linux
Kernel Development Manual, which covers all aspects of the linux-yocto framework
and advanced Linux kernel maintenance concepts.
Summary
In this chapter, we learned how we can customize existing recipes using the
.bbappend files and benefit from this to avoid duplicating source code. We saw how
to enable or disable one feature, how to apply a patch, and how to change BusyBox
and the linux-yocto framework configuration.
In the next chapter, we will discuss how the Yocto Project can help us with some
legal aspects of producing a Linux-based system using packages under different
licenses. We will understand which artifacts we need and how Poky can be
configured to generate the needed artifacts that should be shared as part of the
copyleft compliance accomplishment process.
[ 102 ]
Achieving GPL Compliance
In this chapter, we learn how to fulfill open source license compliance and how we
can use Poky to provide the needed artifacts such as source code, licensing text,
and the list of derivative work. This is critical for most products being introduced
in the market nowadays as open source code needs to live side-by-side with
proprietary code.
Understanding copyleft
Copyleft is a legal way to use the copyright law in order to maximize rights and
express freedom. It impacts our day-to-day work in such a way that companies must
know how to deal with open source and free software licenses as they have a high
impact on their products.
When building a Linux distribution, there are at least two projects being used: the
Linux kernel and a compiler. The most commonly used compiler nowadays is the
GNU Compiler Collection (GCC).
The Linux kernel is released under the GPLv2 license, and the GCC is released under
the GPLv2, GPLv2.1, and GPLv3 licenses, depending on the project used.
However, a Linux-based system can virtually include all projects available
throughout the world, in addition to all applications made by the company for its
product. How do we know the number of projects and licenses, and how do we
fulfill copyleft compliance requirements?
This chapter describes how the Yocto Project can help you in the task,
but be aware that you must know exactly what you need to provide
and the possible license incompatibilities. If you have any doubts,
please consult your legal department or a copyright lawyer.
Achieving GPL Compliance
In this chapter, we understand how the Yocto Project can help us with the most
common tasks required for copyleft compliance.
Copyleft compliance versus proprietary code
It is important to understand that proprietary code and copyleft-covered code can
coexist in the same product. We need to be careful about the libraries we link the
code with because some may have license compatibility issues. However, this is the
standard in most of the available products in the market nowadays.
Some guidelines for license compliance
As already mentioned, one Linux-based system is a set of several projects, each one
under a different license. The Yocto Project helps developers understand that most
copyleft project obligations have the following conditions:
• The source code of the project must be provided along with the binary
• The license of the project must be provided along with the binary
• Any modification in the project or any script needed to configure and build it
must be provided along with the binary
It means that if, one project under copyleft is modified, the license text, the base
source code, and any modification must be included into the final deliverable.
The assumptions cover most rights guaranteed by copyleft licenses. These are the
parts where the Yocto Project may help us. However, before releasing anything,
we are recommended to audit all the materials to be released to make sure
they're complete.
Managing software licensing with Poky
One important Poky feature is the capability of license for license management. Most
of the time, we, as developers, do not care about licenses because we keep our focus
on our own bugs. However, when creating a product, it is very important to care and
know about licenses and the kind of licenses present in it.
Poky keeps track of licenses, works with commercial and noncommercial licenses,
and has a strategy to work with proprietary applications, at least during the
development cycle.
[ 104 ]
Chapter 13
One important thing to know, at first, is that a recipe is released under
a certain license, and it represents a project released under a different
license. The recipe and the project are two different entities and
they have different licensing, so the two different licenses must be
considered part of the product.
In most recipes, information is a comment with the copyright, license, and author
name; this information is regarding the recipe itself. Then, there is a set of variables
to describe the package license, and they are as follows:
• LICENSE: This describes the license under which the package was released.
• LIC_FILES_CHKSUM: This may not seem very useful at first sight. It describes
the license file and its checksum for a package, and we may find a lot of
variation on how a project describes its license. The most common license
files are stored in meta/files/common-licenses/.
Some projects include a file, such as COPYING or LICENSE; others use a header note
on each file or on the main file. The variable LIC_FILES_CHKSUM has the checksum
for the license text of a project; if any letters are changed, the checksum is changed as
well. This is used to make sure any change is noted and consciously accepted by the
developer. A license change may be a typo fix; however, it may also be a change in
legal obligations. So, it is important for the developer to review and understand
the change.
When a different license checksum is detected, BitBake launches a build
error and points to the project that had its license changed. You must
be careful when this happens as the license change may impact the
use of this software. In order to be able to build anything again, you
must change the LIC_FILE_CHKSUM value accordingly and update
the LICENSE field to match the license change. Your legal department
should be consulted if the license terms have changed.
Commercial licenses
By default, Poky does not install any package with a commercial license restriction.
The most used example is the gst-plugins-ugly package. This limitation is
archived though a variable used on these recipes with some license restriction; the
LICENSE_FLAGS variable is used to determinate the restriction.
In the gst-plugins-ugly case, the variable in the recipe is set to LICENSE_FLAGS
= "commercial", although it may have a string. Some projects choose to set it to
LICENSE_FLAGS = "<license>_${PN}_${PV}".
[ 105 ]
Achieving GPL Compliance
In order to install these recipes, we must place a whitelist of desired special licensing
on build/conf/local.conf, and we can do this using LICENSE_FLAGS_WHITELIST.
This variable determines the special license that can be used and has a very
flexible content.
For example, for GStreamer Ugly Plug-ins, we may only want this package to be
installed, so we add the following variable in build/conf/local.conf, as shown:
LICENSE_FLAGS_WHITELIST = "commercial_gst-plugins-ugly"
This exclude any other commercial recipe such as gst-plugins-bad. However, if
we want BitBake to install any commercial package from our image, we may use the
following code in build/conf/local.conf:
LICENSE_FLAGS_WHITELIST = "commercial"
Using Poky to achieve copyleft
compliance
At this point, we know how to use Poky and understand its main goal. It is time to
understand the legal aspects of producing a Linux-based system that uses packages
under different licenses.
We can configure Poky to generate the needed artifacts that should be shared as part
of the copyleft compliance accomplishment process.
License auditing
To help us to achieve copyleft compliance, Poky generates a license manifest during
the image build located at build/tmp/deploy/licenses/<image_name-machine_
name-datestamp>/.
To demonstrate this process, we use the core-image-full-cmdline image for the
qemuarm machine. To start with our example, examine the files under build/tmp/
deploy/licenses/core-image-full-cmdline-qemuarm-<datastamp>, which are
as follows:
• package.manifest: This lists all the packages into the image.
• license.manifest: This lists the names, versions, recipe names, and licenses
for all packages. This file may be used for copyleft compliance auditing.
[ 106 ]
Chapter 13
Providing the source code
The first obvious way Poky may help us to provide the source code of every project
used on our image is by sharing the DL_DIR content. However, this approach has one
important pitfall. Any proprietary source code will be shared within DL_DIR if it is
shared as is. In addition, this approach will share any source code, including the ones
that are not required by copyleft compliance.
Another way is to configure Poky to generate the set of source code and decide
what will be delivered. It may be done using the Archiver class. This class copies
the source code for each package under the build/tmp/deploy folder separated by
architecture (in our example, the present architectures are allarch-poky-linux,
arm-poky-linux-gnueabi, and x86_64-linux) and license. The package for armpoky-linux-gnueabi released under GPLv3 is placed in the build/tmp/deploy/
sources/arm-poky-linux-gnueabi/GPLv3/package-name directory.
Poky must be configured to archive the source code before the final image is
created. So, in order to have it, we can paste the following variables in build/conf/
local.conf:
INHERIT += "archiver"
ARCHIVER_MODE[src] = "original"
Keep in mind that, even with this approach, if we share the build/tmp/deploy/
sources directories, the proprietary unneeded sources are shared, although we may
now choose to share the source based on licensing. We can copy only the packages
under GPLv3 or MIT to a shareable place, or to any other combination of licensing,
according to the desired sharing strategy.
One example of a copy command line, used to copy all packages under any GPL
license (from the Yocto Project Development Manual) is the following:
$: cd poky/build/tmp/deploy/sources
$: mkdir ~/gpl_source_release
$: for dir in */*GPL*; do cp -r $dir ~/gpl_source_release; done
If we prefer to take some help from Poky regarding which license must have our
attention, we can add the ARCHIVER_MODE[filter] ?= "yes" code to build/conf/
local.conf. The default configuration is to have source code for every project,
in other words, no filter. However, if we prefer to have only the source code for
COPYLEFT_LICENSE_INCLUDE projects, we can use a filter.
[ 107 ]
Achieving GPL Compliance
The COPYLEFT_LICENSE_INCLUDE variable currently includes all licenses starting
with GPL or LGPL. This variable can be overridden in build/conf/local.conf, if
we wish to make sure to include another license or variation.
Providing compilation scripts and source
code modifications
With the configuration provided in the previous section, Poky will package the
original source code for each project. In case we want to include the patched source
code, we only use ARCHIVER_MODE[src] = "patched"; this way, Poky will wrap the
project source code after the do_patch task. It includes modifications from recipes or
the bbappend file.
This way, the source code and any modification may be shared easily. However,
there is still one kind of information not created so far: the procedure used to
configure and build the project.
In order to have a reproducible build environment, we may share the configured
project; in other words, the project after the do_configure task. For this, we can add
the following to build/conf/local.conf:
ARCHIVER_MODE[src] = "configured"
It is important to remember that we must consider the person on the other side may
not use the Yocto Project for copyleft compliance; alternatively, if they are using
it, they must know that the modification made on the original source code and
configuration procedure is not available. This is the reason to share the configured
project. It allows anyone to reproduce our build environment.
For all flavors of source code, the default resultant file is a tarball; other options will
add ARCHIVER_MODE[srpm] = "1" in build/conf/local.conf, and the resultant
file will be a SRPM package.
Providing license text
When providing the source code, the license text is shared inside it. If we want
the license text inside our final image, we can add the following to build/conf/
local.conf:
COPY_LIC_MANIFEST = "1"
COPY_LIC_DIRS = "1"
This way, the license files will be placed inside the root filesystem, under /usr/
share/common-licenses/.
[ 108 ]
Chapter 13
Summary
In this chapter, we learned how Poky can help with copyleft license compliance
accomplishment and also understood why it should not be used as a legal
background. Poky enables us to generate source code, reproduction scripts, and
license text for used packages, to be used in our distribution. In addition, we learned
that the license manifest generated within the image may be used to audit the image.
In the next chapter, we will understand how we can use the Yocto Project tools with
an external BSP, the Freescale ARM BSP. We will use it to generate an image for use
with the Wandboard machine.
[ 109 ]
Booting Our
Custom Embedded Linux
It's time! We are now ready to boot our custom-made embedded Linux as we have
learned the required concepts and gained enough knowledge about the Yocto Project
and Poky. In this chapter, we practice what we have learned so far about using Poky
with an external BSP, the Freescale ARM BSP, use it to generate an image for use
with the Wandboard machine, and boot it using the SD card.
The same concepts can be applied to every other board, as long as a vendor provides
a BSP layer to use with the Yocto Project.
We can see a list of the most commonly used BSP layers in this chapter. This should
not be taken as a complete list, or as a definitive one, but we want to facilitate your
search for the needed layer in case you have one board of a specific vendor next to
you. This list is as follows, in alphabetic order:
• Allwinner: This has the meta-allwinner layer
• BeagleBoard: This has the meta-beagleboard layer
• CuBox-i: This has the meta-fsl-arm-extra layer
• Intel: This has the meta-intel layer
• Raspberry Pi: This has the meta-raspberrypi layer
• Texas Instruments: This has the meta-ti layer
• Wandboard: This has the meta-fsl-arm-extra layer
Booting Our Custom Embedded Linux
Exploring the Wandboard
The Wandboard is a not-for-profit low-cost Cortex-A9 processor, based on the
Freescale i.MX6 SoC board with high-performance multimedia capabilities. The
board is available in three versions, as shown in the following table:
Board version
Features
Wandboard Solo
Freescale i.MX6S processor (single-core)
512 MB RAM
Wandboard Dual
Freescale i.MX6DL processor (dual-core)
1 GB RAM
802.11n wireless and Bluetooth
Wandboard Quad
Freescale i.MX6Q processor (quad-core)
2 GB RAM
802.11n wireless and Bluetooth
SATA
The Wandboard is supported by the Wandboard community. More information is
available at http://www.wandboard.org/. The following is an image of a Wandboard:
[ 112 ]
Chapter 14
Discovering Freescale ARM BSP
The Wandboard board uses the Freescale i.MX6 SoC. To use it with the Yocto Project,
we need to use BSP layers.
The needed BSP layers to enable Wandboard support in Yocto Project are
the following:
• meta-fsl-arm: This BSP layer adds support to Freescale's reference machines
and provides the basic BSP support
• meta-fsl-arm-extra: This BSP layer adds support to third-party boards
based on Freescale's SoC, such as Wandboard, and requires meta-fsl-arm as
it provides the base BSP support
Freescale BSP is supported by a community, and more information on it can be
found at http://freescale.github.io/.
Using Wandboard with the Yocto Project
The modular structure of the Yocto Project gives us the freedom to include external
BSP layers to extend a set of supported machines.
The first step to enable support for Wandboard is to download the metadata of the
BSP layers.
From the directory where the Poky source code is cloned, please run the following
command lines:
$: git clone --branch daisy https://github.com/Freescale/meta-fsl-armextra.git
$: git clone --branch daisy git://git.yoctoproject.org/meta-fsl-arm
The final directory structure you should have is shown in the following figure:
sources /
meta-fsl-arm
meta-fsl-arm-extra
poky
[ 113 ]
Booting Our Custom Embedded Linux
After completing this, we must create the build directory we use for our builds. We
can do this using the following command line:
$: source poky/oe-init-build-env build-wandboard
The same procedure was first introduced in Chapter 2, Baking
Our Poky-based System.
Some packages included in Freescale ARM BSP have proprietary property and are
followed by a End-user License Agreement (EULA) that shows the legal impact of
using it. Mainly, the GPU drivers, VPU/IPU codecs, and the meta-fsl-arm layer
have an EULA file that describes the rights and obligations to use the binaries and
source. Read more on the EULA and in case you accept it, edit the build/conf/
local.conf file in order to set ACCEPT_FSL_EULA to 1, as shown in the following line
of code:
ACCEPT_FSL_EULA = "1"
This is not required for the board to work, but for full use, the hardware features are
indispensable.
Independently, whether or not we accept the EULA, we must edit the buildwandboard/conf/bblayers.conf file to enable the just cloned layers. The final file
must look as follows:
[ 114 ]
Chapter 14
The procedure for adding an external layer was first introduced in
Chapter 10, Exploring External Layers.
Building an image for Wandboard
After we have the build directory and the BSP layers are properly set up, we can
start the build. Inside the build-wandboard directory, we must call the
following command:
$: MACHINE=wandboard-<variant> bitbake <image>
The MACHINE variable can be changed depending on the Wandboard we want to use
or set in build/conf/local.conf. The machine names for the Wandboard variants
are wandboard-solo, wandboard-dual, and wandboard-quad.
If we want to use the Wandboard Solo and build core-image-sato, which provides
an embedded graphical environment, we should run the following command:
$: MACHINE=wandboard-solo bitbake core-image-sato
The build process will take a while. It takes about 2 hours and about 10 Gigabytes in
a workstation machine.
If you didn't accept the EULA or want a smaller image to build faster, you can build
core-image-minimal instead. For this, use the following command:
$: MACHINE=wandboard-solo bitbake core-image-minimal
In Chapter 2, Baking Our Poky-based System, there is a list of
some possible images to be used.
Booting Wandboard from the SD card
After the build process is over, the image will be available inside the buildwandboard/tmp/deploy/images/wandboard-solo/ directory. There are many files,
but Freescale ARM BSP generates a ready-to-use SD card image.
The file we want to use is core-image-sato-wandboard-solo.sdcard or coreimage-minimal.sdcard, depending on the image we built.
[ 115 ]
Booting Our Custom Embedded Linux
Make sure your point to the right device and double check to not
write in your hard disk.
In order to copy the generated image to the SD card, we should use the dd utility,
as follows:
$: sudo dd if=core-image-sato-wandboard-solo.sdcard of=/dev/sdX bs=1M
We can also use the following command:
$: sudo dd if=core-image-minimal-wandboard-solo.sdcard of=/dev/sdX
bs=1M
After copying the content in the SD card, insert it into the SD card slot, connect the
HDMI cable, and power on the machine. It should boot nicely.
There are two SD card slots in Wandboard. The primary slot is
located in the CPU board, used for booting, and a secondary slot is
found in the peripheral board (the base board).
Summary
In this final chapter, we introduced two community projects, Wandboard and
Freescale ARM BSP layers. We consolidated our Yocto Project knowledge by adding
external BSP layers and using these in a real board with a generated image.
Throughout the book, we learned the needed background information for you to
learn on your own any other aspect of the Yocto Project that you may need. You have
the general understanding to know what is happening behind the scenes when you
ask BitBake to build a recipe or an image. From now on, you are ready to free your
mind and try new things. There are plenty of boards available, waiting for you to
give them life. The ball is in your court now; here's where the fun begins!
[ 116 ]
References
The following are the references used in the book:
• Yocto Project Reference Manual: http://www.yoctoproject.org/
docs/1.6/ref-manual/ref-manual.html
• BitBake Development mailing list:
bitbake-devel@lists.openembedded.org
• OpenEmbedded-Core mailing list: openembedded-core@lists.
openembedded.org
• Yocto Project Development Manual: http://www.yoctoproject.org/
docs/1.6/dev-manual/dev-manual.html
• Yocto Project Git: http://git.yoctoproject.org/
• OpenEmbedded Layer Index: http://layers.openembedded.org
• BitBake User Manual: http://www.yoctoproject.org/docs/1.6/
bitbake-user-manual/bitbake-user-manual.html
• Yocto Project Linux Kernel Development Manual:
http://www.yoctoproject.org/docs/1.6/kernel-dev/kernel-dev.html
Index
Symbols
:= operator 61
??= operator
used, for setting default value of
metadata 61
?= operator
used, for setting default value of
metadata 61
.= operator 62
+= operator 62
=.operator 62
=+ operator 62
A
appending, metadata 62
application debugging
differentiating, with metadata
debugging 73
ARM Versatile Platform Baseboard 44
Autoconf based recipes
options, adding to 98
Autotools 91
B
basic variable, metadata
setting 60
BitBake
about 8, 9, 27
tasks 35, 36
used, for modifying kernel
configuration 102
BitBake metadata. See metadata
bitbake-prserv command 56
BitBake tasks
building 42
configuration 42
fetching 42
installing 42
package, creating 42
source preparation 42
sysroot, wrapping 42
Board Support Package (BSP) 16
board versions, Wandboard
Wandboard Dual 112
Wandboard Quad 112
Wandboard Solo 112
BSP layers, Yocto Project
Allwinner 111
BeagleBoard 111
CuBox-i 111
Intel 111
Raspberry Pi 111
Texas Instruments 111
Wandboard 111
build-appliance-image 17
build directory
about 41
conf 41
constructing 42
downloads 41
sstate-cache 41
tmp 41
build environment
setting up 15, 16
buildhistory
used, for tracking image 74, 75
used, for tracking packages 74, 75
used, for tracking SDK 74, 75
buildhistory-diff utility 74
build-time dependency 29
BusyBox
about 97, 101
customizing 101
C
classes 28
CMake 91
commercial licenses 105
conditional appending, metadata 63
conditional metadata 63
conf directory 41
configuration files 28
configuration, host system 13, 14
copyleft 103
copyleft compliance
versus proprietary code 104
copyleft compliance, achieving
with Poky
about 106
compilation scripts, providing 108
license auditing 106
license text, providing 108
source code modifications, providing 108
source code, providing 107
Cortex-A9 processor 112
custom application
developing, on target 70, 71
integrating, with Eclipse 72
custom distribution
using 93-95
customization, BusyBox 101
customization, linux-yocto
framework 101, 102
custom layers
creating 87, 88
metadata, adding to 89
D
dash shell 64
DEB 38
Debian
about 14
Poky, installing on 14
Debian Package Manager 50
DEB package format 50
debugging
about 73
application debugging 73
metadata debugging 73
with GDB 78
debugging, packages 75, 76
dependency
about 29, 30
build-time dependency 29
runtime dependency 29
DEPENDS variable 29
deploy directory 43
development shell. See devshell
devshell
about 73
using 77
distribution layer 83
DISTRO_FEATURES variable
versus, MACHINE_FEATURES variable 95
DISTRO_FEATURES variable, supported
values
alsa 94
bluetooth 94
cramfs 94
ext2 94
ipsec 94
ipv6 94
irda 94
keyboard 95
nfs 95
opengl 95
pci 95
pcmcia 95
ppp 95
smbfs 95
systemd 95
usbgadget 95
usbhost 95
wayland 95
wifi 95
do_compile task 36
do_configure task 36
do_fetch task 36
do_install_append function 99
do_install task 36
[ 120 ]
do_package task 36
do_patch task 36
Dora 54
do_unpack task 36
downloads directory 41
Dropbear 90
E
Eclipse
about 72
custom application, integrating with 72
End-user License Agreement (EULA) 114
executable metadata
defining 64
F
Fedora
about 14
Poky, installing on 14
fetcher backend 31
file inclusion, metadata 63
files
adding, to existing packages 99
file searching paths 99, 100
framebuffer 94
Freescale ARM BSP
about 113
meta-fsl-arm 113
meta-fsl-arm-extra 113
G
GDB
about 78
used, for debugging 78
generic SDK 69
Gentoo 27
Gentoo Portage package system 10
Git
URL 15
Git repository 31-33
GNU Compiler Collection (GCC) 103
GNU Make 27
GNU Project Debugger. See GDB
GPLv2.1 license 103
GPLv2 license 103
GPLv3 license 103
GStreamer Ugly Plug-ins 106
H
Hob
about 21
used, for building image 21-23
used, for customizing image 23-26
host system
configuring 13, 14
Poky, installing on Debian 14
Poky, installing on Fedora 14
I
image
building, for Wandboard 115
building, Hob used 21-23
creating 89, 90
customizing, Hob used 23-26
tracking, buildhistory used 74, 75
image-based SDK
using 68-70
IMAGE_FEATURES variable
about 57
dbg-pkgs 90
dev-pkgs 90
doc-pkgs 90
nfs-server 90
read-only-rootfs 90
splash 90
ssh-server-dropbear 90
ssh-server-openssh 90
staticdev-pkgs 90
tools-debug 90
tools-profile 90
tools-sdk 91
tools-testapps 91
x11 91
x11-base 91
x11-sato 91
immediate variable expansion,
metadata 61, 62
include directive 63
inheritance system, metadata 65
[ 121 ]
inherit directive 65
installation, Poky
on Debian 14
on Fedora 14
IPK 38, 50
Itsy Package Management System 50
K
kernel configuration
modifying, BitBake used 102
modifying, SDK used 102
L
LAYERDEPENDS variable 88
layers
about 81-83
Board Support Package (BSP) 82
custom layers, creating 87, 88
distribution layer 83
meta layers, adding 85, 86
software layer 82
source code, detailing 83, 84
layers, properties
Name property 82
Path property 82
Priority property 82
LAYERVERSION variable 88
license compliance
guidelines 104
LICENSE_FLAGS variable 105
LICENSE variable, package license 105
LIC_FILES_CHKSUM variable, package
license 105
linux-yocto framework
customizing 101, 102
Linux Yocto Framework 97
local.conf file 16, 17
logging functions, Python
bb.debug 76
bb.error 76
bb.fatal 76
bb.note 76
bb.plain 76
bb.warn 76
logging functions, Shell Script
bbdebug 76
bberror 76
bbfatal 76
bbnote 76
bbplain 76
bbwarn 76
M
machine definition
creating 92, 93
IMAGE_FSTYPES variable 93
KERNEL_IMAGETYPE variable 93
MACHINE_FEATURES variable 92
PREFERRED_PROVIDER_virtual/kernel
variable 92
SERIAL_CONSOLES variable 92
TARGET_ARCH variable 92
MACHINE_FEATURES variable
acpi 93
alsa 93
apm 93
bluetooth 93
ext2 93
irda 93
keyboard 93
pci 93
pcmcia 93
screen 93
serial 93
touchscreen 93
usbgadget 93
usbhost 93
versus, DISTRO_FEATURES variable 95
wifi 93
MACHINE variable 16
metadata
about 9, 10
adding, to custom layers 89
appending 62
basic variable, setting 60
classes 28
conditional appending 63
conditional metadata 63
configuration files 28
default value, setting with ??= operator 61
[ 122 ]
default value, setting with ?= operator 61
executable metadata, defining 64
exploring 28
file inclusion 63
immediate variable expansion 61, 62
inheritance system 65
meta-yocto 10
meta-yocto-bsp 10
parsing 28, 29
prepending 62
Python functions, defining 64, 65
Python variable expansion 64
recipes 28
using 59
variable expansion 60
working with 60
metadata, classification
classes 59
configuration 59
recipes 59
metadata debugging
differentiating, with application
debugging 73
meta-fsl-arm, BSP layers 113
meta-fsl-arm-extra, BSP layers 113
meta layers
adding 85, 86
URL 85
meta-toolchain 69
meta-yocto 10
meta-yocto-bsp 10, 16
MIRRORS variable 33, 34
mklibs process 38
N
network access
disabling 35
disabling, reasons 35
O
OpenedHand Sato 91
OpenEmbedded-Core 9
OpenEmbedded project 10
OpenEmbedded Project and Yocto
Project alliance 10
OpenMoko 50
OpenSSH source code 31
OpenWrt 50
OpenZaurus project 10
Opkg fork 50
options
adding, to Autoconf based recipes 98
overrides 63
P
PACKAGE_CLASSES variable 50
package feed
about 55, 56
using 56, 57
PACKAGE_FEED_URIS variable 56
package recipe
adding 91
packages
debugging 75, 76
files, adding to 99
patch, applying over 98
tracking, buildhistory used 74, 75
package versioning 54
patch
applying, over packages 98
PE variable 54
Poky
about 8
BitBake, using 9
installing, on Debian 14
installing, on Fedora 14
metadata 10
objective 9
OpenEmbedded-Core metadata
collection 9
used, for managing software
licensing 104, 105
poky directory 15
Poky SDK
about 68
generic SDK 69
image-based SDK, using 68-70
Poky source code
downloading 14, 15
Portage 27
postinst script 51
[ 123 ]
postrm script 51
preinst script 51
prelink process 38
PREMIRRORS variable 33, 34
prepending, metadata 62
prerm script 51
PR variable 54
psplash 90
Python functions, metadata
defining 64, 65
Python variable expansion, metadata 64
Q
QEMU
about 19
target image, executing in 19, 20
qemuarm directory 46
qemuarm-tcbootstrap directory 46
QMake 91
qt4e-demo-image 18
Quick EMUlator. See QEMU
R
RDEPENDS variable 29
recipe feature configuration
modifying 100
recipes
about 28
preferring 30, 31
providing 30, 31
Red Hat Package Manager 49
remote file downloads 31, 32
require directive 63
root filesystem image
generating 37-39
rootfs image. See root filesystem image
RPM 38, 49
runqemu script 19
runtime dependency 29
S
SDK
about 67, 68
tracking, buildhistory used 74, 75
used, for modifying kernel
configuration 102
shared state cache 49, 53
smart install command 57
smart query command 57
software development kit. See SDK
software layer 82
software licensing
managing, Poky used 104, 105
source code
fetching 31
Git repository 32, 33
network access, disabling 35
other repositories 33
remote file downloads 31, 32
source code download
optimizing 33, 34
source code, layers
detailing 83, 84
SRCREV variable 33
sstate-cache directory 41
sstate-cache package 53
supported package formats
code, running 51
DEB 50
IPK 50
RPM 49
scripts, using 51, 52
selecting 50
TAR 50
using 49
sysroots directory
about 43, 46
qemuarm directory 46
qemuarm-tcbootstrap directory 46
x86_64-linux directory 46
T
target image
build-appliance-image 17
building 17-19
core-image-base 17
core-image-clutter 18
core-image-directfb 18
core-image-full-cmdline 18
core-image-lsb 18
[ 124 ]
core-image-lsb-dev 18
core-image-lsb-sdk 18
core-image-minimal 17
core-image-minimal-dev 17
core-image-minimal-initramfs 17
core-image-minimal-mtdutils 18
core-image-multilib-example 18
core-image-rt 18
core-image-rt-sdk 18
core-image-sato 18
core-image-sato-dev 18
core-image-sato-sdk 18
core-image-weston 18
core-image-x11 18
executing, in QEMU 19, 20
qt4e-demo-image 18
TAR package format 50
tasks
about 35, 36
do_compile task 36
do_configure task 36
do_fetch task 36
do_install task 36
do_package task 36
do_patch task 36
do_unpack task 36
extending 37
temporary build directory
deploy 43
exploring 42
work 43
TFTP server 101
tmp directory 41
toolchain 67
V
variable expansion, metadata 60
variables scope 96
W
Wandboard
about 112
board versions 112
booting, from SD card 115
image, building 115
URL 112
using, with Yocto Project 113, 114
work directory
<sources> 44
about 43-45
image 44
packages 44
packages-split 44
temp 44
X
x86_64-linux directory 46
Y
Yocto Project
about 7
delineating 8
online references 117
Poky 8
U
use cases 97
[ 125 ]
Thank you for buying
Embedded Linux Development with Yocto Project
About Packt Publishing
Packt, pronounced 'packed', published its first book "Mastering phpMyAdmin for Effective
MySQL Management" in April 2004 and subsequently continued to specialize in publishing
highly focused books on specific technologies and solutions.
Our books and publications share the experiences of your fellow IT professionals in adapting
and customizing today's systems, applications, and frameworks. Our solution based books
give you the knowledge and power to customize the software and technologies you're using
to get the job done. Packt books are more specific and less general than the IT books you have
seen in the past. Our unique business model allows us to bring you more focused information,
giving you more of what you need to know, and less of what you don't.
Packt is a modern, yet unique publishing company, which focuses on producing quality,
cutting-edge books for communities of developers, administrators, and newbies alike.
For more information, please visit our website: www.packtpub.com.
About Packt Open Source
In 2010, Packt launched two new brands, Packt Open Source and Packt Enterprise, in order
to continue its focus on specialization. This book is part of the Packt Open Source brand,
home to books published on software built around Open Source licenses, and offering
information to anybody from advanced developers to budding web designers. The Open
Source brand also runs Packt's Open Source Royalty Scheme, by which Packt gives a royalty
to each Open Source project about whose software a book is sold.
Writing for Packt
We welcome all inquiries from people who are interested in authoring. Book proposals should
be sent to author@packtpub.com. If your book idea is still at an early stage and you would like
to discuss it first before writing a formal book proposal, contact us; one of our commissioning
editors will get in touch with you.
We're not just looking for published authors; if you have strong technical skills but no writing
experience, our experienced editors can help you develop a writing career, or simply get some
additional reward for your expertise.
Instant Optimizing Embedded
Systems Using BusyBox
ISBN: 978-1-78328-985-1
Paperback: 72 pages
Learn to optimize embedded systems with BusyBox
using practical, hands-on recipes
1.
Build and play embedded (Android Linux)
system with BusyBox from scratch. Configure,
compile, cross-compile, and install BusyBox.
2.
Build external utilities and development
environments (include Bash and C) for a
BusyBox-based embedded system to meet
diverse system requirements.
Linux Utilities Cookbook
ISBN: 978-1-78216-300-8
Paperback: 224 pages
Over 70 recipes to help you accomplish a wide
variety of tasks in Linux quickly and efficiently
1.
Use the command line like a pro.
2.
Pick a suitable desktop environment.
3.
Learn to use files and directories efficiently.
Please check www.PacktPub.com for information on our titles
BeagleBone Home Automation
ISBN: 978-1-78328-573-0
Paperback: 178 pages
Live your sophisticated dream with home automation
using BeagleBone
1.
Practical approach to home automation using
BeagleBone; starting from the very basics of
GPIO control and progressing up to building a
complete home automation solution.
2.
Covers the operating principles of a range of
useful environment sensors, including their
programming and integration to the server
application.
Kali Linux – Assuring Security by
Penetration Testing
ISBN: 978-1-84951-948-9
Paperback: 454 pages
Master the art of penetration testing with Kali Linux
1.
Learn penetration testing techniques with an
in-depth coverage of Kali Linux distribution.
2.
Explore the insights and importance of testing
your corporate network systems before the
hackers strike.
3.
Understand the practical spectrum of security
tools by their exemplary usage, configuration,
and benefits.
Please check www.PacktPub.com for information on our titles