Interrupts

Introduction

Imagine you are reading a book (executing a program) and someone taps your shoulder (interrupt). You mark your page (save state), address the interruption (run the handler), then resume reading from where you left off (restore state). This is exactly how CPU interrupts work — they allow external events to get immediate attention without the CPU constantly checking for them.

Interrupts are the nervous system of a computer. They allow hardware devices to signal the CPU that they need attention, the OS to handle errors and system calls, and real-time events to be processed promptly. Without interrupts, the CPU would have to constantly poll every device, wasting enormous amounts of processing power.

Types of Interrupts

Hardware Interrupts (External)

Generated by hardware devices:

• Keyboard key pressed
• Disk read completed
• Network packet arrived
• Timer tick (for scheduling)
• Mouse moved

Software Interrupts (Traps)

Generated by programs intentionally:

• System calls (INT 0x80 on Linux, SYSCALL instruction)
• Breakpoints (for debuggers)
• Division by zero
• Page faults
• Invalid memory access (segfault)

The Interrupt Handling Process

Interrupt Vector Table (IVT)

A table mapping interrupt numbers to handler addresses:

Interrupt Priority and Masking

Not all interrupts are equally urgent. A priority system ensures critical interrupts are handled first:

• Non-Maskable Interrupts (NMI): Cannot be disabled. Reserved for critical errors (hardware failure, memory parity error).
• Maskable Interrupts: Can be temporarily disabled. Most device interrupts fall here. The OS disables them during critical kernel sections to prevent data corruption.

Interrupt-Driven I/O

Instead of the CPU continuously checking if a device is ready (polling), interrupt-driven I/O lets the CPU do other work and be notified when the device is ready:

Context of Interrupts in Modern Systems

Modern CPUs handle thousands of interrupts per second. A typical desktop system processes:

• ~1000 timer interrupts/second (for scheduling)
• Variable keyboard/mouse interrupts
• Hundreds to thousands of disk/network interrupts during active I/O

Each interrupt costs ~1-10 microseconds, so at 10,000 interrupts/second, the overhead is ~1-10% of CPU time.

Real-World Analogy

Interrupts are like notifications on your phone. Without them, you would have to manually check every app every few seconds (polling) to see if anything new happened. With interrupts (notifications), you can focus on your current task and be alerted only when something needs attention. Priority levels are like notification settings — calls ring loud (NMI), messages vibrate (normal interrupt), and social media is silenced (masked).

Key Takeaways

• Interrupts allow devices to signal the CPU asynchronously without CPU polling
• Hardware interrupts come from devices; software interrupts (traps) come from programs
• The IVT maps interrupt numbers to handler routine addresses
• Interrupt handling saves state, runs the handler, then restores state transparently
• Priority and masking control which interrupts can preempt which
• Interrupt-driven I/O is fundamental to efficient multitasking — the CPU works while waiting for I/O
• Modern systems handle thousands of interrupts per second with microsecond overhead each