Why is Linux so pervasive? And why does something as simple as a TV need to run something as complex as Linux just to display streaming video on a screen?

The simple answer is Moore's Law: Gordon Moore, co-founder of Intel, observed in 1965 that the density of components on a chip will double approximately every 2 years. That applies to the devices that we design and use in our everyday lives just as much as it does to desktops, laptops, and servers. At the heart of most embedded devices is a highly integrated chip that contains one or more processor cores and interfaces with main memory, mass storage, and peripherals of many types. This is referred to as a System on Chip, or SoC, and SoCs are increasing in complexity in accordance with Moore's Law. A typical SoC has a technical reference manual that stretches to thousands of pages. Your TV is not simply displaying a video stream as the old analog sets used to do.

The stream is digital, possibly encrypted, and it needs processing to create an image. Your TV is (or soon will be) connected to the internet. It can receive content from smartphones, tablets, and home media servers. It can be (or soon will be) used to play games and so on. You need a full operating system to manage this degree of complexity.

Here are some points that drive the adoption of Linux:

• Linux has the necessary functionality. It has a good scheduler, a good network stack, support for USB, Wi-Fi, Bluetooth, many kinds of storage media, good support for multimedia devices, and so on. It ticks all the boxes.
• Linux has been ported to a wide range of processor architectures, including some that are very commonly found in SoC designs – Arm, MIPS, x86, and PowerPC.
• Linux is open source, so you have the freedom to get the source code and modify it to meet your needs. You, or someone working on your behalf, can create a board support package for your particular SoC board or device. You can add protocols, features, and technologies that may be missing from the mainline source code. You can remove features that you don't need to reduce memory and storage requirements. Linux is flexible.
• Linux has an active community; in the case of the Linux kernel, very active. There is a new release of the kernel every 8 to 10 weeks, and each release contains code from more than 1,000 developers. An active community means that Linux is up to date and supports current hardware, protocols, and standards.
• Open source licenses guarantee that you have access to the source code. There is no vendor tie-in.

For these reasons, Linux is an ideal choice for complex devices. But there are a few caveats I should mention here. Complexity makes it harder to understand. Coupled with the fast-moving development process and the decentralized structures of open source, you have to put some effort into learning how to use it and to keep on re-learning as it changes. I hope that this book will help in the process.

Is Linux suitable for your project? Linux works well where the problem being solved justifies the complexity. It is especially good where connectivity, robustness, and complex user interfaces are required. However, it cannot solve every problem, so here are some things to consider before you jump in:

• Is your hardware up to the job? Compared to a traditional real-time operating system (RTOS) such as VxWorks or QNX, Linux requires a lot more resources. It needs at least a 32-bit processor and lots more memory. I will go into more detail in the section on typical hardware requirements.
• Do you have the right skill set? The early parts of a project, board bring-up, require detailed knowledge of Linux and how it relates to your hardware. Likewise, when debugging and tuning your application, you will need to be able to interpret the results. If you don't have the skills in-house, you may want to outsource some of the work. Of course, reading this book helps!
• Is your system real-time? Linux can handle many real-time activities so long as
  you pay attention to certain details, which I will cover in detail in Chapter 21,
  Real-Time Programming.
• Will your code require regulatory approval (medical, automotive, aerospace, and so on)? The burden of regulatory verification and validation might make another OS a better choice. Even if you do choose Linux for use in these environments, it may make sense to purchase a commercially available distribution from a company that has supplied Linux for existing products, like the one you are building.

Consider these points carefully. Probably the best indicator of success is to look around for similar products that run Linux and see how they have done it; follow best practice.

Where does open source software come from? Who writes it? In particular, how does this relate to the key components of embedded development—the toolchain, bootloader, kernel, and basic utilities found in the root filesystem?

The main players are as follows:

• The open source community: This, after all, is the engine that generates the software you are going to be using. The community is a loose alliance of developers, many of whom are funded in some way, perhaps by a not-for-profit organization, an academic institution, or a commercial company. They work together to further the aims of the various projects. There are many of them—some small, some large. Some that we will be making use of in the remainder of this book are Linux itself, U-Boot, BusyBox, Buildroot, the Yocto Project, and the many projects under the GNU umbrella.
• CPU architects: These are the organizations that design the CPUs we use. The important ones here are Arm/Linaro (Arm Cortex-A), Intel (x86 and x86_64), SiFive (RISC-V), and IBM (PowerPC). They implement or, at the very least, influence support for the basic CPU architecture.
• SoC vendors (Broadcom, Intel, Microchip, NXP, Qualcomm, TI, and many others): They take the kernel and toolchain from the CPU architects and modify them to support their chips. They also create reference boards: designs that are used by the next level down to create development boards and working products.
• Board vendors and OEMs: These people take the reference designs from SoC vendors and build them in to specific products, for instance, set-top boxes or cameras, or create more general-purpose development boards, such as those from Advantech and Kontron. An important category are the cheap development boards such as BeagleBoard/BeagleBone and Raspberry Pi that have created their own ecosystems of software and hardware add-ons.
• Commercial Linux vendors: Companies such as Siemens (Mentor), Timesys, and Wind River offer commercial Linux distributions that have undergone strict regulatory verification and validation across multiple industries (medical, automotive, aerospace, and so on).

These form a chain, with your project usually at the end, which means that you do not have a free choice of components. You cannot simply take the latest kernel from https://www.kernel.org/, except in a few rare cases, because it does not have support for the chip or board that you are using.

This is an ongoing problem with embedded development. Ideally, the developers at each link in the chain would push their changes upstream, but they don't. It is not uncommon to find a kernel that has many thousands of patches that are not merged. In addition, SoC vendors tend to actively develop open source components only for their latest chips, meaning that support for any chip more than a couple of years old will be frozen and not receive any updates.

The consequence is that most embedded designs are based on old versions of software. They do not receive security fixes, performance enhancements, or features that are in newer versions. Problems such as Heartbleed (a bug in the OpenSSL libraries) and ShellShock (a bug in the bash shell) go unfixed. I will talk more about this later in this chapter under the topic of security.

What can you do about it? First, ask questions of your vendors (NXP, Texas Instruments, and Xilinx, to name just a few): what is their update policy, how often do they revise kernel versions, what is the current kernel version, what was the one before that, and what is their policy for merging changes upstream? Some vendors are making great strides in this way. You should prefer their chips.

Secondly, you can take steps to make yourself more self-sufficient. The chapters in Section 1 explain the dependencies in more detail and show you where you can help yourself. Don't just take the package offered to you by the SoC or board vendor and use it blindly without considering the alternatives.

This book is divided into four sections that reflect the phases of a project. The phases are not necessarily sequential. Usually, they overlap and you will need to jump back to revisit things that were done previously. However, they are representative of a developer's preoccupations as the project progresses:

• Elements of Embedded Linux (Chapters 1 to 8) will help you set up the development environment and create a working platform for the later phases. It is often referred to as the board bring-up phase.
• System Architecture and Design Choices (Chapters 9 to 15) will help you to look at some of the design decisions you will have to make concerning the storage of programs and data, how to divide work between kernel device drivers and applications, and how to initialize the system.
• Writing Embedded Applications (Chapters 16 to 18) shows how to package and deploy Python applications, make effective use of the Linux process and thread model, and how to manage memory in a resource-constrained device.
• Debugging and Optimizing Performance (Chapters 19 to 21) describes how to trace, profile, and debug your code in both the applications and the kernel. The last chapter explains how to design for real-time behavior when required.

Now, let's focus on the four basic elements of embedded Linux that comprise the first section of the book.

The four elements of embedded Linux

Every project begins by obtaining, customizing, and deploying these four elements: the toolchain, the bootloader, the kernel, and the root filesystem. This is the topic of the first section of this book.

• Toolchain: The compiler and other tools needed to create code for your
  target device.
• Bootloader: The program that initializes the board and loads the Linux kernel.
• Kernel: This is the heart of the system, managing system resources and interfacing with hardware.
• Root filesystem: Contains the libraries and programs that are run once the kernel has completed its initialization.

Of course, there is also a fifth element, not mentioned here. That is the collection of programs specific to your embedded application that make the device do whatever it is supposed to do, be it weigh groceries, display movies, control a robot, or fly a drone.

Typically, you will be offered some or all of these elements as a package when you buy your SoC or board. But, for the reasons mentioned in the preceding paragraph, they may not be the best choices for you. I will give you the background to make the right selections in the first eight chapters and I will introduce you to two tools that automate the whole process for you: Buildroot and the Yocto Project.

The components of embedded Linux are open source, so now is a good time to consider what that means, why open sources work the way they do, and how this affects the often proprietary embedded device you will be creating from it.

Licenses

When talking about open source, the word free is often used. People new to the subject often take it to mean nothing to pay, and open source software licenses do indeed guarantee that you can use the software to develop and deploy systems for no charge. However, the more important meaning here is freedom, since you are free to obtain the source code, modify it in any way you see fit, and redeploy it in other systems. These licenses give you this right. Compare that with freeware licenses, which allow you to copy the binaries for no cost but do not give you the source code, or other licenses that allow you to use the software for free under certain circumstances, for example, for personal use, but not commercial. These are not open source.

I will provide the following comments in the interest of helping you understand the implications of working with open source licenses, but I would like to point out that I am an engineer and not a lawyer. What follows is my understanding of the licenses and the way they are interpreted.

Open source licenses fall broadly into two categories: copyleft licenses, such as the GNU General Public License (GPL), and permissive licenses, such as the BSD and MIT licenses.

The permissive licenses say, in essence, that you may modify the source code and use it in systems of your own choosing so long as you do not modify the terms of the license in any way. In other words, with that one restriction, you can do with it what you want, including building it into possibly proprietary systems.

The GPL licenses are similar but have clauses that compel you to pass the rights to obtain and modify the software on to your end users. In other words, you share your source code. One option is to make it completely public by putting it onto a public server. Another is to offer it only to your end users by means of a written offer to provide the code when requested. The GPL goes further to say that you cannot incorporate GPL code into proprietary programs. Any attempt to do so would make the GPL apply to the whole. In other words, you cannot combine a GPL and proprietary code in one program. Aside from the Linux kernel, the GNU Compiler Collection and GNU Debugger as well as many other freely available tools associated with the GNU project fall under the umbrella of the GPL.

So, what about libraries? If they are licensed with the GPL, any program linked with them becomes GPL also. However, most libraries are licensed under the GNU Lesser General Public License (LGPL). If this is the case, you are allowed to link with them from a proprietary program.

Important note

All of the preceding description relates specifically to GPL v2 and LGPL v2.1. I should mention the latest versions of GPL v3 and LGPL v3. These are controversial, and I will admit that I don't fully understand the implications. However, the intention is to ensure that the GPL v3 and LGPL v3 components in any system can be replaced by the end user, which is in the spirit of open source software for everyone.

The GPL v3 and LGPL v3 have their problems though. There are issues with security. If the owner of a device has access to the system code, then so might an unwelcome intruder. Often the defense is to have kernel images that are signed by an authority such as the vendor, so that unauthorized updates are not possible. Is that an infringement of my right to modify my device? Opinions differ.

Important note

The TiVo set-top box is an important part of this debate. It uses a Linux kernel, which is licensed under GPL v2. TiVo have released the source code of their version of the kernel and so comply with the license. TiVo also has a bootloader that will only load a kernel binary that is signed by them. Consequently, you can build a modified kernel for a TiVo box, but you cannot load it on the hardware. The Free Software Foundation (FSF) takes the position that this is not in the spirit of open source software and refers to this procedure as Tivoization. The GPL v3 and LGPL v3 were written to explicitly prevent this from happening. Some projects, the Linux kernel in particular, have been reluctant to adopt the GPL version 3 licenses because of the restrictions they would place on device manufacturers.

If you are designing or selecting hardware for an embedded Linux project, what do you look out for?

First, a CPU architecture that is supported by the kernel—unless you plan to add a new architecture yourself, of course! Looking at the source code for Linux 5.4, there are 25 architectures, each represented by a sub-directory in the arch/ directory. They are all
32- or 64-bit architectures, most with an MMU, but some without. The ones most often found in embedded devices are Arm, MIPS, PowerPC, and x86, each in 32 and 64-bit variants, all of which have memory management units (MMUs).

Most of this book is written with this class of processor in mind. There is another group that doesn't have an MMU and that runs a subset of Linux known as microcontroller Linux or uClinux. These processor architectures include ARC (Argonaut RISC Core), Blackfin, MicroBlaze, and Nios. I will mention uClinux from time to time, but I will not go into detail because it is a rather specialized topic.

Second, you will need a reasonable amount of RAM. 16 MiB is a good minimum, although it is quite possible to run Linux using half that. It is even possible to run Linux with 4 MiB if you are prepared to go to the trouble of optimizing every part of the system. It may even be possible to get lower, but there comes a point at which it is no longer Linux.

Third, there is non-volatile storage, usually flash memory. 8 MiB is enough for a simple device such as a webcam or a simple router. As with RAM, you can create a workable Linux system with less storage if you really want to, but the lower you go, the harder it becomes. Linux has extensive support for flash storage devices, including raw NOR and NAND flash chips, and managed flash in the form of SD cards, eMMC chips, USB flash memory, and so on.

Fourth, a serial port is very useful, preferably a UART-based serial port. It does not have to be fitted on production boards, but makes board bring-up, debugging, and development much easier.

Fifth, you need some means of loading software when starting from scratch. Many microcontroller boards are fitted with a Joint Test Action Group (JTAG) interface for this purpose. Modern SoCs also have the ability to load boot code directly from removable media, especially SD and micro SD cards, or serial interfaces such as UART or USB.

In addition to these basics, there are interfaces to the specific bits of hardware your device needs to get its job done. Mainline Linux comes with open source drivers for many thousands of different devices, and there are drivers (of variable quality) from the SoC manufacturer and from the OEMs of third-party chips that may be included in the design, but remember my comments on the commitment and ability of some manufacturers. As a developer of embedded devices, you will find that you spend quite a lot of time evaluating and adapting third-party code, if you have it, or liaising with the manufacturer if you don't. Finally, you will have to write the device support for interfaces that are unique to the device or find someone to do it for you.

The examples in this book are intended to be generic, but to make them relevant and easy to follow, I have had to choose specific hardware. I have chosen three exemplar devices: the Raspberry Pi 4, the BeagleBone Black, and QEMU. The first is by far the most popular Arm-based single board computer on the market. The second is a widely available and cheap development board that can be used in serious embedded hardware. The third is a machine emulator that can be used to create a range of systems that are typical of embedded hardware. It was tempting to use QEMU exclusively, but, like all emulations, it is not quite the same as the real thing. Using the Raspberry Pi 4 and BeagleBone Black, you have the satisfaction of interacting with real hardware and seeing real LEDs flash. While the BeagleBone Black is several years old now, it remains open source hardware (unlike the Raspberry Pi). This means that the board design materials are freely available for anyone to build a BeagleBone Black or derivative into their products.

In any case, I encourage you to try out as many of the examples as you can, using either of these three platforms, or indeed any embedded hardware you may have to hand.

The Raspberry Pi 4

At the time of writing, the Raspberry Pi 4 Model B is the flagship tiny, dual-display, desktop computer produced by the Raspberry Pi Foundation. Their website is
https://raspberrypi.org/. The Pi 4's technical specs include the following:

• A Broadcom BCM2711 1.5 GHz quad-core Cortex-A72 (Arm® v8) 64-bit SoC
• 2, 4, or 8 GiB DDR4 RAM
• 2.4 GHz and 5.0 GHz 802.11ac wireless, Bluetooth 5.0, BLE
• A serial port for debug and development
• A MicroSD slot, which can be used as the boot device
• A USB-C connector that is used to power the board
• 2 × full size USB 3.0 and 2 × full size USB 2.0 host ports
• A Gigabit Ethernet port
• 2 × micro-HDMI ports for video and audio output

In addition, there is a 40-pin expansion header for which there are a great variety of daughter boards, known as HATs (Hardware Attached on Top), that allow you to adapt the board to do many different things. However, you will not need any HATs for the examples in this book. Instead, you will make use of the Pi 4's built-in Wi-Fi and Bluetooth (which the BeagleBone Black lacks).

In addition to the board itself, you will require the following:

• A 5V USB-C power supply capable of delivering 3 A or more
• A USB to TTL serial cable with 3.3V logic-level pins like the Adafruit 954
• A MicroSD card and a means of writing to it from your development PC or laptop, which will be needed to load software onto the board
• An Ethernet cable and a router to connect it to, as some of the examples require network connectivity

Next is the BeagleBone Black.

The BeagleBone Black

The BeagleBone and the later BeagleBone Black are open hardware designs for a small, credit card-sized development board produced by CircuitCo LLC. The main repository of information is at https://beagleboard.org/. The main points of the specifications are as follows:

• A TI AM335x 1 GHz Arm® Cortex-A8 Sitara SoC
• 512 MiB DDR3 RAM
• 2 or 4 GiB 8-bit eMMC onboard flash storage
• A serial port for debug and development
• A MicroSD slot, which can be used as the boot device
• A Mini-USB OTG client/host port that can also be used to power the board
• A full size USB 2.0 host port
• A 10/100 Ethernet port
• An HDMI port for video and audio output

In addition, there are two 46-pin expansion headers for which there are a great variety of daughter boards, known as capes, which allow you to adapt the board to do many different things. However, you do not need to fit any capes for the examples in this book.

In addition to the board itself, you will require the following:

• A Mini-USB to USB-A cable (supplied with the board).
• A serial cable that can interface with the 6-pin 3.3V TTL level signals provided by the board. The BeagleBoard website has links to compatible cables.
• A MicroSD card and a means of writing to it from your development PC or laptop, which will be needed to load software onto the board.
• An Ethernet cable and a router to connect it to, as some of the examples require network connectivity.
• A 5V power supply capable of delivering 1 A or more.

In addition to the above, Chapter 12, Prototyping with Breakout Boards, also requires
the following:

• A SparkFun model GPS-15193 Breakout board.
• A Saleae Logic 8 logic analyzer. This apparatus will be used to probe pins for SPI communications between the BeagleBone Black and NEO-M9N.

QEMU

QEMU is a machine emulator. It comes in a number of different flavors, each of which can emulate a processor architecture and a number of boards built using that architecture. For example, we have the following:

• qemu-system-arm: 32-bit Arm
• qemu-system-mips: MIPS
• qemu-system-ppc: PowerPC
• qemu-system-x86: x86 and x86_64

For each architecture, QEMU emulates a range of hardware, which you can see by using the -machine help option. Each machine emulates most of the hardware that would normally be found on that board. There are options to link hardware to local resources, such as using a local file for the emulated disk drive.

Here is a concrete example:

$ qemu-system-arm -machine vexpress-a9 -m 256M -drive file=rootfs.ext4,sd -net nic -net use -kernel zImage -dtb vexpress- v2p-ca9.dtb -append "console=ttyAMA0,115200 root=/dev/mmcblk0" -serial stdio -net nic,model=lan9118 -net tap,ifname=tap0

The options used in the preceding command line are as follows:

• -machine vexpress-a9: Creates an emulation of an Arm Versatile Express development board with a Cortex A-9 processor
• -m 256M: Populates it with 256 MiB of RAM
• -drive file=rootfs.ext4,sd: Connects the SD interface to the local file rootfs.ext4 (which contains a filesystem image)
• -kernel zImage: Loads the Linux kernel from the local file named zImage
• -dtb vexpress-v2p- ca9.dtb: Loads the device tree from the local file vexpress-v2p-ca9.dtb
• -append "...": Appends this string as the kernel command line
• -serial stdio: Connects the serial port to the terminal that launched QEMU, usually so that you can log on to the emulated machine via the serial console
• -net nic,model=lan9118: Creates a network interface
• -net tap,ifname=tap0: Connects the network interface to the virtual network interface, tap0

To configure the host side of the network, you need the tunctl command from the User Mode Linux (UML) project; on Debian and Ubuntu, the package is named uml-utilites:

$ sudo tunctl -u $(whoami) -t tap0

This creates a network interface named tap0 that is connected to the network controller in the emulated QEMU machine. You configure tap0 in exactly the same way as any other interface.

All of these options are described in detail in the following chapters. I will be using Versatile Express for most of my examples, but it should be easy to use a different machine or architecture.

I have only used open source software, both for the development tools and the target operating system and applications. I assume that you will be using Linux on your development system. I tested all the host commands using Ubuntu 20.04 LTS, and so there is a slight bias toward that particular version, but any modern Linux distribution is likely to work just fine.

Embedded hardware will continue to get more complex, following the trajectory set by Moore's Law. Linux has the power and the flexibility to make use of hardware in an efficient way. Together we will learn how to harness that power so we can build robust products that delight our users. This book will take you through the five phases of the embedded project's life cycle, beginning with the four elements of embedded Linux.

The sheer variety of embedded platforms and the fast pace of development lead to isolated pools of software. In many cases, you will become dependent on this software, especially the Linux kernel that is provided by your SoC or board vendor, and, to a lesser extent, the toolchain. Some SoC manufacturers are getting better at pushing their changes upstream and the maintenance of these changes is getting easier. Despite these improvements, selecting the right hardware for your embedded Linux project is still an exercise fraught with peril. Open source license compliance is another topic you need to be aware when building products atop the embedded Linux ecosystem.

In this chapter, you were introduced to the hardware and some of the software you will use throughout this book (namely QEMU). Later on, we will examine some powerful tools that can help you create and maintain the software for your device. We cover Buildroot and dig deep into the Yocto Project. Before I describe these build tools, I will describe the four elements of embedded Linux, which you can apply to all embedded Linux projects, however they are created.

The next chapter is all about the first of these, the toolchain, which you need in order to compile code for your target platform.