⚙️ Operating System

An operating system (OS) is a layer of abstraction between software and hardware in a computer. Its primary job is to manage 💼 Processes on the system, ensuring that they’re isolated from each other and providing an illusion of exclusive access to the processor. It also protects the system by serving as a middle-man between the running process (which can be malicious) and the hardware.

Kernel §

The kernel of an OS is the memory-resident OS code that’s used for important functions such as context switching, running the 🗓️ Scheduler, and handling ⏰ Interrupts. It has the power to run certain CPU instructions (direct device I/O, page table manipulation) and acts as a layer of safety between custom user programs and the hardware.

System Calls §

For user programs to use privileged instructions, it must interact with the kernel via a system call. Externally, they look just like normal function calls with arguments and return values; internally, they save the values in registers (rather than stack frame) and execute kernel code. The steps are as follows:

1. Load arguments (small ones into registers, big ones into memory).
2. Load system call number into register.
3. Execute trap instruction to enter kernel mode, jump to pre-defined kernel “entry point” address.
4. Kernel saves process’s context into its PCB.
5. Kernel jumps to system call handler.
6. System call handler runs. If this handler blocks (for example, due to I/O), the process is blocked and we switch to another one via the 🗓️ Scheduler.
7. System call handler loads return values.
8. Kernel restores process context from PCB, executes return instruction.

Multiprocessing §

In a basic hardware system, a single CPU reads and executes instructions sequentially (known as control flow). Thus, we cannot run multiple programs at once.

However, multiprocessing gives the illusion of multiple simultaneously-running programs by interleaving the execution of the programs. These processes are called “concurrent”—control flows overlapping in time, with one flow running at any specific moment. Note that this is slightly different from parallel processes, where there can be multiple flows running at the same moment in time (requiring multiple processors).

Control Flow §

Control flow is the sequence of instructions a processor executes. Individual programs can control the flow with jumps, branches, calls, and returns, but they only respond to program state. Changes to the system state—I/O, timers, networking, and more—also change the control flow, either with ⏰ Interrupts, context switches, or 🗼 Signals.

Context Switch §

A context switch passes control from one process to another. Each time this happens, user code stops, kernel code responsible for the context switch runs (save old process state in old PCB, switch address space, load new state from new PCB), and then the new process’s user code runs. Note that this process is somewhat expensive, and its overhead needs to be taken into account.