Introduction to Kernel Development
Linux started as a hobby project in 1991 for a Finnish student, Linus Torvalds. The project has gradually grown and still does, with roughly 1,000 contributors around the world. Nowadays, Linux is a must, in embedded systems as well as on servers. A kernel is a center part of an operating system, and its development is not so obvious.
Linux offers many advantages over other operating systems:
- It is free of charge
- Well documented with a large community
- Portable across different platforms
- Provides access to the source code
- Lots of free open source software
This book tries to be as generic as possible. There is a special topic, device tree, which is not a full x86 feature yet. That topic will then be dedicated to ARM processors, and all those fully supporting the device tree. Why those architectures? Because they are most used on desktop and servers (for x86), and on embedded systems (ARM).
This chapter deals, among other things, with:
- Development environment setup
- Getting, configuring, and building kernel sources
- Kernel source code organization
- Introduction to kernel coding style
Environment setup
Before you start any development, you need to set an environment up. The environment dedicated to Linux development is quite simple, at least on Debian-based systems:
$ sudo apt-get update $ sudo apt-get install gawk wget git diffstat unzip texinfo \ gcc-multilib build-essential chrpath socat libsdl1.2-dev \ xterm ncurses-dev lzop
There are parts of code in this book that are compatible with ARM system on chip (SoC). You should install gcc-arm as well:
sudo apt-get install gcc-arm-linux-gnueabihf
I'm running Ubuntu 16.04, on an ASUS RoG, with an Intel core i7 (eight physical cores), 16 GB of RAM, 256 GB of SSD, and 1 TB of magnetic hard drive. My favorite editor is Vim, but you are free to use the one you are most comfortable with.
Getting the sources
In the early kernel days (until 2003), odd-even versioning styles were used, where odd numbers were stable and even numbers were unstable. When the 2.6 version was released, the versioning scheme switched to X.Y.Z, where:
- X: This was the actual kernel version, also called major; it incremented when there were backwards-incompatible API changes
- Y: This was the minor revision; it incremented after adding a functionality in a backwards-compatible manner
- Z: This is also called PATCH, representing the version relative to bug fixes
This is called semantic versioning, and has been used until the 2.6.39 version; when Linus Torvalds decided to bump the version to 3.0, which also meant the end of semantic versioning in 2011, and then an X.Y scheme was adopted.
When it came to the 3.20 version, Linus argued that he could no longer increase Y, and decided to switch to an arbitrary versioning scheme, incrementing X whenever Y got large enough that he ran out of fingers and toes to count it. This is the reason why the version has moved from 3.20 to 4.0 directly. Have a look at https://plus.google.com/+LinusTorvalds/posts/jmtzzLiiejc.
Now, the kernel uses an arbitrary X.Y versioning scheme, which has nothing to do with semantic versioning.
Source organization
For the needs of this book, you must use Linus Torvald's GitHub repository:
git clone https://github.com/torvalds/linux git checkout v4.1 ls
- arch/: The Linux kernel is a fast growing project that supports more and more architectures. That being said, the kernel wants to be as generic as possible. Architecture-specific code is separated from the rest, and falls into this directory. This directory contains processor-specific subdirectories such as alpha/, arm/, mips/, blackfin/, and so on.
- block/: This directory contains code for block storage devices, actually the scheduling algorithm.
- crypto/: This directory contains the cryptographic API and the encryption algorithms code.
- Documentation/: This should be your favorite directory. It contains the descriptions of APIs used for different kernel frameworks and subsystems. You should look here prior to asking any questions on forums.
- drivers/: This is the heaviest directory, continuously growing as device drivers get merged. It contains every device driver organized in various subdirectories.
- fs/: This directory contains the implementation of different filesystems that the kernel actually supports, such as NTFS, FAT, ETX{2,3,4}, sysfs, procfs, NFS, and so on.
- include/: This contains kernel header files.
- init/: This directory contains the initialization and start up code.
- ipc/: This contains implementation of the Inter-Process Communication (IPC) mechanisms, such as message queues, semaphores, and shared memory..
- kernel/: This directory contains architecture-independent portions of the base kernel.
- lib/: Library routines and some helper functions live here. They are generic kernel object (kobject) handlers, Cyclic Redundancy Code (CRC) computation functions, and so on.
- mm/: This contains memory management code.
- net/: This contains networking (whatever network type it is) protocols code.
- scripts/: This contains scripts and tools used during kernel development. There are other useful tools here.
- security/: This directory contains the security framework code.
- sound/: Audio subsystems code is here.
- usr/: This currently contains the initramfs implementation.
The kernel must remain portable. Any architecture-specific code should be located in the arch directory. Of course, the kernel code related to the user space API does not change (system calls, /proc, /sys), as it would break the existing programs.
Kernel configuration
The Linux kernel is a makefile-based project, with thousands of options and drivers. To configure your kernel, either use make menuconfig for an ncurse-based interface or make xconfig for an X-based interface. Once chosen, options will be stored in a .config file, at the root of the source tree.
In most cases, there will be no need to start a configuration from scratch. There are default and useful configuration files available in each arch directory, which you can use as a starting point:
ls arch/<you_arch>/configs/
For ARM-based CPUs, these configs files are located in arch/arm/configs/, and for an i.MX6 processor, the default file config is arch/arm/configs/imx_v6_v7_defconfig. Similarly, for x86 processors we find the files in arch/x86/configs/, with only two default configuration files, i386_defconfig and x86_64_defconfig, for 32- and 64-bit versions respectively. It is quite straightforward for an x86 system:
make x86_64_defconfig make zImage -j16 make modules makeINSTALL_MOD_PATH </where/to/install> modules_install
Given an i.MX6-based board, you can start with ARCH=arm make imx_v6_v7_defconfig, and then ARCH=arm make menuconfig. With the former command, you will store the default option in the .config file, and with the latter, you can update add/remove options, depending on the need.
You may run into a Qt4 error with xconfig. In such a case, you should just use the following command:
sudo apt-get install qt4-dev-tools qt4-qmake
Building your kernel
Building the kernel requires you to specify the architecture for which it is built, as well as the compiler. That said, it is not necessary for a native build:
ARCH=arm make imx_v6_v7_defconfig ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- make zImage -j16
After that, you will see something like:
[...] LZO arch/arm/boot/compressed/piggy_data CC arch/arm/boot/compressed/misc.o CC arch/arm/boot/compressed/decompress.o CC arch/arm/boot/compressed/string.o SHIPPED arch/arm/boot/compressed/hyp-stub.S SHIPPED arch/arm/boot/compressed/lib1funcs.S SHIPPED arch/arm/boot/compressed/ashldi3.S SHIPPED arch/arm/boot/compressed/bswapsdi2.S AS arch/arm/boot/compressed/hyp-stub.o AS arch/arm/boot/compressed/lib1funcs.o AS arch/arm/boot/compressed/ashldi3.o AS arch/arm/boot/compressed/bswapsdi2.o AS arch/arm/boot/compressed/piggy.o LD arch/arm/boot/compressed/vmlinux OBJCOPY arch/arm/boot/zImage Kernel: arch/arm/boot/zImage is ready
From the kernel build, the result will be a single binary image located in arch/arm/boot/. Modules are built with the following command:
ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- make modules
You can install them using the following command:
ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- make modules_install
The modules_install target expects an environment variable, INSTALL_MOD_PATH, which specifies where you should install the modules. If not set, the modules will be installed at /lib/modules/$(KERNELRELEASE)/kernel/. This is discussed in Chapter 2, Device Driver Basis.
i.MX6 processors support device trees, which are files you use to describe the hardware (this is discussed in detail in Chapter 6, The Concept of Device Tree). To compile every ARCH device tree, you can run the following command:
ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- make dtbs
However, the dtbs option is not available on all platforms that support device tree. To build a standalone DTB, you should use:
ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- make imx6d- sabrelite.dtb
Kernel habits
The kernel code tried to follow standard rules throughout its evolution. In this chapter, we will just be introduced to them. They are all discussed in a dedicated chapter; starting from Chapter 3, Kernel Facilities and Helper Functions, we get a better overview of the kernel development process and tips, up to Chapter 13, Linux Device Model.
Coding style
Before going deep into this section, you should always refer to the kernel coding style manual, at Documentation/CodingStyle in the kernel source tree. This coding style is a set of rules you should respect, at least if you need to get patches accepted by kernel developers. Some of these rules concern indentation, program flow, naming conventions, and so on.
The most popular ones are:
- Always use a tab indentation of eight characters, and the line should be 80 columns long. If the indentation prevents you from writing your function, it is because this one has too many nesting levels. One can size the tabs and verify the line size using a scripts/cleanfile script from the kernel source:
scripts/cleanfile my_module.c
- You can also indent the code correctly using the indent tool:
sudo apt-get install indent scripts/Lindent my_module.c
- Every function/variable that is not exported should be declared as static.
- No spaces should be added around (inside) parenthesized expressions. s = size of (struct file); is accepted, whereas s = size of( struct file ); is not.
- Using typdefs is forbidden.
- Always use /* this */ comment style, not // this:
- BAD: // do not use this please
- GOOD: /* Kernel developers like this */
- You should capitalise macros, but functional macros can be in lowercase.
- A comment should not replace code that is not illegible. Prefer rewriting the code rather than adding a comment.
Kernel structure allocation/initialization
The kernel always offers two possible allocation mechanisms for its data structures and facilities.
Some of these structures are:
- Workqueue
- List
- Waitqueue
- Tasklet
- Timer
- Completion
- mutex
- spinlock
Dynamical initializers are all macros, which means they are always capitalized: INIT_LIST_HEAD(), DECLARE_WAIT_QUEUE_HEAD(), DECLARE_TASKLET( ), and so on.
These are all discussed in Chapter 3, Kernel Facilities and Helper Functions. Data structures that represent framework devices are always allocated dynamically, each having its own allocation and deallocation API. These framework device types are:
- Network
- Input device
- Char device
- IIO device
- Class
- Framebuffer
- Regulator
- PWM device
- RTC
The scope of the static objects is visible in the whole driver, and by every device this driver manages. Dynamically allocated objects are visible only by the device that is actually using a given instance of the module.
Classes, objects, and OOP
The kernel implements OOP by means of a device and a class. Kernel subsystems are abstracted by means of classes. There are almost as many subsystems as there are directories under /sys/class/. The struct kobject structure is the centerpiece of this implementation. It even brings in a reference counter, so that the kernel may know how many users actually use the object. Every object has a parent, and has an entry in sysfs (if mounted).
Every device that falls into a given subsystem has a pointer to an operations (ops) structure, which exposes operations that can be executed on this device.
Summary
This chapter explained in a very short and simple manner how you should download the Linux source and process a first build. It also dealt with some common concepts. That said, this chapter is quite brief and may not be enough, but never mind, it is just an introduction. That is why the next chapter gets more into the details of the kernel building process; how to actually compile a driver, either externally or as a part of the kernel; as well as some basics that you should learn before starting the long journey that kernel development represents.