A program is a set of computer instructions that perform a particular task.
That program can be written in assembler, a very low level computer language,
or in a high level, machine independent language such as the C programming
language.
An operating system is a special program which allows the user to run applications such as
spreadsheets and word processors.
This chapter introduces basic programming principles and gives an overview of
the aims and functions of an operating system.
The instructions that a CPU fetches from memory and executes are not at all understandable to human beings. They are machine codes which tell the computer precisely what to do. The hexadecimal number 0x89E5 is an Intel 80486 instruction which copies the contents of the ESP register to the EBP register. One of the first software tools invented for the earliest computers was an assembler, a program which takes a human readable source file and assembles it into machine code. Assembly languages explicitly handle registers and operations on data and they are specific to a particular microprocessor. The assembly language for an Intel X86 microprocessor is very different to the assembly language for an Alpha AXP microprocessor. The following Alpha AXP assembly code shows the sort of operations that a program can perform:
ldr r16, (r15) ; Line 1
ldr r17, 4(r15) ; Line 2
beq r16,r17,100 ; Line 3
str r17, (r15) ; Line 4
100: ; Line 5
The first statement (on line 1) loads register 16 from the address held in register 15. The next instruction loads register 17 from the next location in memory. Line 3 compares the contents of register 16 with that of register 17 and, if they are equal, branches to label 100. If the registers do not contain the same value then the program continues to line 4 where the contents of r17 are saved into memory. If the registers do contain the same value then no data needs to be saved. Assembly level programs are tedious and tricky to write and prone to errors. Very little of the Linux kernel is written in assembly language and those parts that are are written only for efficiency and they are specific to particular microprocessors.
Writing large programs in assembly language is a difficult and time consuming task. It is prone to error and the resulting program is not portable, being tied to one particular processor family. It is far better to use a machine independent language like C. C allows you to describe programs in terms of their logical algorithms and the data that they operate on. Special programs called compilers read the C program and translate it into assembly language, generating machine specific code from it. A good compiler can generate assembly instructions that are very nearly as efficient as those written by a good assembly programmer. Most of the Linux kernel is written in the C language. The following C fragment:
if (x != y)
x = y ;
performs exactly the same operations as the previous example assembly code. If the contents of the variable x are not the same as the contents of variable y then the contents of y will be copied to x. C code is organized into routines, each of which perform a task. Routines may return any value or data type supported by C. Large programs like the Linux kernel comprise many separate C source modules each with its own routines and data structures. These C source code modules group together logical functions such as filesystem handling code.
C supports many types of variables, a variable is a location in memory which can be referenced by a symbolic name. In the above C fragment x and y refer to locations in memory. The programmer does not care where in memory the variables are put, it is the linker (see below) that has to worry about that. Some variables contain different sorts of data, integer and floating point and others are pointers.
Pointers are variables that contain the address, the location in memory of other data. Consider a variable called x, it might live in memory at address 0x80010000. You could have a pointer, called px, which points at x. px might live at address 0x80010030. The value of px would be 0x80010000: the address of the variable x.
C allows you to bundle together related variables into data structures. For example,
struct {
int i ;
char b ;
} my_struct ;
is a data structure called my_struct which contains two elements, an integer (32 bits of data storage) called i and a character (8 bits of data) called b.
Linkers are programs that link together several object modules and libraries to form a single, coherent, program. Object modules are the machine code output from an assembler or compiler and contain executable machine code and data together with information that allows the linker to combine the modules together to form a program. For example one module might contain all of a program's database functions and another module its command line argument handling functions. Linkers fix up references between these object modules, where a routine or data structure referenced in one module actually exists in another module. The Linux kernel is a single, large program linked together from its many constituent object modules.
Without software a computer is just a pile of electronics that gives off heat. If the hardware is the heart of a computer then the software is its soul. An operating system is a collection of system programs which allow the user to run application software. The operating system abstracts the real hardware of the system and presents the system's users and its applications with a virtual machine. In a very real sense the software provides the character of the system. Most PCs can run one or more operating systems and each one can have a very different look and feel. Linux is made up of a number of functionally separate pieces that, together, comprise the operating system. One obvious part of Linux is the kernel itself; but even that would be useless without libraries or shells.
In order to start understanding what an operating system is, consider what happens when you type an apparently simple command:
$ ls Mail c images perl docs tcl $
The $ is a prompt put out by a login shell (in this case bash). This means that it is waiting for you, the user, to type some command. Typing ls causes the keyboard driver to recognize that characters have been typed. The keyboard driver passes them to the shell which processes that command by looking for an executable image of the same name. It finds that image, in /bin/ls. Kernel services are called to pull the ls executable image into virtual memory and start executing it. The ls image makes calls to the file subsystem of the kernel to find out what files are available. The filesystem might make use of cached filesystem information or use the disk device driver to read this information from the disk. It might even cause a network driver to exchange information with a remote machine to find out details of remote files that this system has access to (filesystems can be remotely mounted via the Networked File System or NFS). Whichever way the information is located, ls writes that information out and the video driver displays it on the screen.
All of the above seems rather complicated but it shows that even most simple commands reveal that an operating system is in fact a co-operating set of functions that together give you, the user, a coherent view of the system.
With infinite resources, for example memory, many of the things that an operating system has to do would be redundant. One of the basic tricks of any operating system is the ability to make a small amount of physical memory behave like rather more memory. This apparently large memory is known as virtual memory. The idea is that the software running in the system is fooled into believing that it is running in a lot of memory. The system divides the memory into easily handled pages and swaps these pages onto a hard disk as the system runs. The software does not notice because of another trick, multi-processing.
A process could be thought of as a program in action, each process is a separate entity that is running a particular program. If you look at the processes on your Linux system, you will see that there are rather a lot. For example, typing ps shows the following processes on my system:
$ ps PID TTY STAT TIME COMMAND 158 pRe 1 0:00 -bash 174 pRe 1 0:00 sh /usr/X11R6/bin/startx 175 pRe 1 0:00 xinit /usr/X11R6/lib/X11/xinit/xinitrc -- 178 pRe 1 N 0:00 bowman 182 pRe 1 N 0:01 rxvt -geometry 120x35 -fg white -bg black 184 pRe 1 < 0:00 xclock -bg grey -geometry -1500-1500 -padding 0 185 pRe 1 < 0:00 xload -bg grey -geometry -0-0 -label xload 187 pp6 1 9:26 /bin/bash 202 pRe 1 N 0:00 rxvt -geometry 120x35 -fg white -bg black 203 ppc 2 0:00 /bin/bash 1796 pRe 1 N 0:00 rxvt -geometry 120x35 -fg white -bg black 1797 v06 1 0:00 /bin/bash 3056 pp6 3 < 0:02 emacs intro/introduction.tex 3270 pp6 3 0:00 ps $
If my system had many CPUs then each process could (theoretically at least) run on a different CPU. Unfortunately, there is only one so again the operating system resorts to trickery by running each process in turn for a short period. This period of time is known as a time-slice. This trick is known as multi-processing or scheduling and it fools each process into thinking that it is the only process. Processes are protected from one another so that if one process crashes or malfunctions then it will not affect any others. The operating system achieves this by giving each process a separate address space which only they have access to.
Device drivers make up the major part of the Linux kernel. Like other parts of the operating system, they operate in a highly privileged environment and can cause disaster if they get things wrong. Device drivers control the interaction between the operating system and the hardware device that they are controlling. For example, the filesystem makes use of a general block device interface when writing blocks to an IDE disk. The driver takes care of the details and makes device specific things happen. Device drivers are specific to the controller chip that they are driving which is why, for example, you need the NCR810 SCSI driver if your system has an NCR810 SCSI controller.
In Linux, as it is for Unix TM , the separate filesystems that the system may use are not accessed by device identifiers (such as a drive number or a drive name) but instead they are combined into a single hierarchical tree structure that represents the filesystem as a single entity. Linux adds each new filesystem into this single filesystem tree as they are mounted onto a mount directory, for example /mnt/cdrom. One of the most important features of Linux is its support for many different filesystems. This makes it very flexible and well able to coexist with other operating systems. The most popular filesystem for Linux is the EXT2 filesystem and this is the filesystem supported by most of the Linux distributions.
A filesystem gives the user a sensible view of files and directories held on the hard disks of the system regardless of the filesystem type or the characteristics of the underlying physical device. Linux transparently supports many different filesystems (for example MS-DOS and EXT2) and presents all of the mounted files and filesystems as one integrated virtual filesystem. So, in general, users and processes do not need to know what sort of filesystem that any file is part of, they just use them.
The block device drivers hide the differences between the physical block device types (for example, IDE and SCSI) and, so far as each filesystem is concerned, the physical devices are just linear collections of blocks of data. The block sizes may vary between devices, for example 512 bytes is common for floppy devices whereas 1024 bytes is common for IDE devices and, again, this is hidden from the users of the system. An EXT2 filesystem looks the same no matter what device holds it.
Mostly these data structures exist in physical memory and are accessible only by the kernel and its subsystems. Data structures contain data and pointers; addresses of other data structures or the addresses of routines. Taken all together, the data structures used by the Linux kernel can look very confusing. Every data structure has a purpose and although some are used by several kernel subsystems, they are more simple than they appear at first sight.
Understanding the Linux kernel hinges on understanding its data structures and the use that the various functions within the Linux kernel makes of them. This book bases its description of the Linux kernel on its data structures. It talks about each kernel subsystem in terms of its algorithms, its methods of getting things done, and their usage of the kernel's data structures.
A hash table is an array of pointers to data structures and its index is derived from information in those data structures. If you had data structures describing the population of a village then you could use a person's age as an index. To find a particular person's data you could use their age as an index into the population hash table and then follow the pointer to the data structure containing the person's details. Unfortunately many people in the village are likely to have the same age and so the hash table pointer becomes a pointer to a chain or list of data structures each describing people of the same age. However, searching these shorter chains is still faster than searching all of the data structures.
As a hash table speeds up access to commonly used data structures, Linux often uses hash tables to implement caches. Caches are handy information that needs to be accessed quickly and are usually a subset of the full set of information available. Data structures are put into a cache and kept there because the kernel often accesses them. There is a drawback to caches in that they are more complex to use and maintain than simple linked lists or hash tables. If the data structure can be found in the cache (this is known as a cache hit, then all well and good. If it cannot then all of the relevant data structures must be searched and, if the data structure exists at all, it must be added into the cache. In adding new data structures into the cache an old cache entry may need discarding. Linux must decide which one to discard, the danger being that the discarded data structure may be the next one that Linux needs.
Often these lower layers register themselves with the upper layer at boot time. This registration usually involves adding a data structure to a linked list. For example each filesystem built into the kernel registers itself with the kernel at boot time or, if you are using modules, when the filesystem is first used. You can see which filesystems have registered themselves by looking at the file /proc/filesystems. The registration data structure often includes pointers to functions. These are the addresses of software functions that perform particular tasks. Again, using filesystem registration as an example, the data structure that each filesystem passes to the Linux kernel as it registers includes the address of a filesystem specfic routine which must be called whenever that filesystem is mounted.