Lectures: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13.

Operating System Architecture

The goal of this self study module is to refresh knowledge on basic computer functionality, in particular on the representation of a program in the computer memory, and to look at the basic architecture of the operating system and the work an operating system has to do when loading programs.

From prior courses, rehearsed in this module, you should be able to:

  • describe the instruction cycle of a common processor,
  • give examples of possible processor instructions,
  • describe the use of registers in a processor,
  • demonstrate the use of instructions to build simple program constructs.

At the end of this module, you should be able to:

  • give examples of abstractions provided by the operating system for the applications,
  • describe common operating system architectures (monolithic, microkernel, hypervisor),
  • describe the purpose of an operating system kernel,
  • identify and give examples of privileged operations,
  • use tools to disassemble code implementing simple program constructs,
  • examine and explain the disassembled code for common processor instructions,
  • explain how the disassembled code is produced from the binary program representation,
  • describe alternatives to static compilation from source code to binary program representation (interpretation, just-in-time compilation).

Basic Computer Functionality

The background for this lecture includes awareness of basic computer functionality, such as taught in the NSWI120 Principles of Computers and NSWI170 Computer Systems courses. Please consult the course pages for refresher materials.

At a very basic level, processor is a simple machine that keeps fetching and executing instructions. While knowing all instructions provided by a specific processor is rarely necessary, it is important to have a general idea of what an instruction can look like.

[Q] Consider the following general descriptions of operations:

  1. add constant 123 to value in register X
  2. copy value from register X to register Y
  3. compute sin of value in register X and store result in register Y
  4. read value from memory address 123 and store it in register X
  5. write value from register X to memory variable V
  6. print character with code in register X to screen
  7. copy memory block from address 123 of size 456 to address 789

Which of the operations above could be provided as actual processor instructions in a common processor ?

Hint ...

The key to answering this question is understanding the level of abstraction at which the processor operates. The individual processor instructions operate on registers and memory addressed using numerical addresses. The operations may vary in complexity between RISC and CISC processors.

A textual notation is used to represent processor instructions for human consumption. Some compilers can produce such notation on demand, this is useful when we want to see how particular program constructs translate to processor instructions. Try the following on a Linux computer (or try the Compiler Explorer):

int i = 1234;
int main (void) { return (i); }

With GCC:

gcc -S main.c
less main.s

With CLANG:

clang -S main.c
less main.s

Machine Program Representation

For machine consumption, processor instructions are stored simply as numbers in memory. Specific numbers denote specific instructions and their operands. Encoding of instructions to numbers depend on the particular processor type, examples can be found in the processor manuals.

Check out an example of instruction encoding for Intel in the Intel Processor Manual Volume 2 Appendix A Opcode Map. Check out an example of instruction encoding for MIPS in the MIPS Processor Manual Section 2.1 CPU Instruction Formats. Check out an example of instruction encoding for RISC-V in the RISC-V Processor Manual Volume 1 Section 2.2 Base Instruction Formats.

To see examples of instruction encoding, we can disassemble the binary format produced by the compiler:

With GCC:

gcc -o main main.c
objdump -d main

With CLANG:

clang -o main main.c
objdump -d main

Obviously, the processor will attempt to interpret any memory content as code if told to do so. This is just for fun:

echo 'Hello !' > code.bin
objdump -b binary -m i386:x86-64 -D code.bin

[Q] On modern computers, the examples above would be compiled in 64-bit mode by default. That mode produces somewhat more complicated code than the 32-bit mode. Add the -m32 compiler command line switch to the examples above to get the 32-bit code rather than the 64-bit code.

In the generated code, look at the body of the main function, particularly how it accesses the i variable. How is the address of the i variable specified in the assembly source and in the disassembly ?

Hint ...

The output may differ depending on the compiler used, however, recognizing the i variable in the assembly source should be easier than in the disassembly. Why ?

Processors support various addressing modes. Read about Intel addressing modes in the Intel Processor Manual Volume 1 Section 3.7.5 Specifying an Offset. Read about MIPS addressing modes in the MIPS Processor Manual Appendix A Instruction LW. Read about RISC-V addressing modes in the RISC-V Processor Manual Volume 1 Section 2.6 Load and Store Instructions.

[Q] Look at the code from the example above. What memory addresses is it located at ? What would need to be done if it were to be moved to a different range of addresses ?

Hint ...

Think about how the binary program representation handles addresses.

Operating System Architecture

The role of the operating system is to manage shared resources and provide shared abstractions for the applications executing on the computer. On the management part, the main goals are fair (or otherwise controlled) resource allocation and protection. On the abstraction part, the operating system provides concepts such as processes, sockets, files, and so on.

Resource protection relies on introducing the concept of privileged operations, which only the operating system is permitted to execute. The architecture of the operating system typically reflects this concept by defining the system kernel, which executes the privileged operations, and possibly additional system services, which have no special privileges compared to applications.

Read up on the concept of privileged operations in Arpaci-Dusseau Section 6 Limited Direct Execution. Note that the text uses the term “virtualize” in a broad sense of “provide the illusion of having an own execution environment”.

[Q] Give an example of a (high level) operating system function that needs to execute privileged operations.

Hint ...

Operating system functions are usually exposed to applications in the form of library functions. A common example of such system library is GNU libc, used by many Linux distributions. However, not all library functions require privileged operations.

Since privileged operations can only be executed by the kernel, and the kernel is invoked (among other) through system calls, looking at the list of system calls may provide another hint. Use man syscalls to get such list.

[Q] Give an example of a (low level) program operation that would be considered privileged.

Hint ...

Remember, privileged operations can only be executed by the kernel. By definition then, a program would have to attempt something illegal.

Eager For More ?

Want to delve into the topic beyond the standard module content ? We’ve got you covered !