In-Depth Notes on Main Memory and Addressing in Operating Systems

Main Memory Overview

• Programs must be loaded from disk into memory to be executed.
• The CPU accesses memory via a direct connection with main memory and registers (fast access).
• Main memory is slower; hence, cache memory is used to alleviate latency.

Memory Protection

• Ensures processes access only their allocated address space.
• Achieved using base and limit registers to define logical address space.

Address Binding

• Programs on disk are queued before loading.
• Addresses during different program phases:
  • Source code: symbolic address representation.
  • Compiled code: relocatable addresses (e.g., “14 bytes from the start”).
  • Loader: binds relocatable to absolute addresses.

Instruction and Data Binding to Memory

• Address binding occurs at three stages:
  • Compile time: Absolute code, needs recompilation if location changes.
  • Load time: Relocatable code is generated.
  • Execution time: Binding occurs at runtime, requires hardware support (e.g., MMU with base/limit registers).

Logical vs Physical Address Space

• Logical address: generated by the CPU (virtual address).
• Physical address: directly recognized by memory.
• Mapping occurs via a memory-management unit (MMU).

Memory-Management Unit (MMU)

• Maps virtual addresses to physical addresses during execution.
• Utilizes a relocation register that adjusts all addresses generated by the process before they are sent to memory.

Dynamic Loading and Linking

• Dynamic Loading: Routine loaded only when called.
  • Results in better memory utilization.
• Dynamic Linking: Linking is delayed until execution, using stubs to resolve addresses for library routines.

Contiguous Allocation

• Early method of memory allocation: resident OS in low memory, user processes in high memory.
• Utilizes relocation registers to protect user processes, ensuring each logical address is within bounds.

Variable Partition

• Memory allocated in partitions of varied size, improving efficiency.
• When a process finishes, its partition can be combined with adjacent free partitions.

Dynamic Storage Allocation Problem

• Addressing allocation strategy:
  • First-fit.
  • Best-fit.
  • Worst-fit.
• First-fit and best-fit are typically more efficient.

Fragmentation

• External Fragmentation: Available memory in total but not contiguous.
• Internal Fragmentation: Allocated memory greater than requested.
• Solutions include memory compaction (only feasible with dynamic relocation).

Paging

• Allows noncontiguous physical address space, aiding in reducing external fragmentation.
• Divides memory into fixed-sized blocks called frames.
• Logical memory divided into pages, with a page table facilitating address mapping.
• Pages can suffer from internal fragmentation.

Address Translation in Paging

• Logical address consists of a page number and page offset.
• Physical addresses derived using the page table (base address of each page).

TLB (Translation Lookaside Buffer)

• Cache used for faster address translation.
• Provides significant performance improvements; management policies are crucial.

Memory Protection Techniques

• Protection bits associated with frames for access rights.
• Valid-invalid bits in the page table for region access control.

Shared Pages

• Efficient for reentrant code shared among processes, each process retains private data copies.

Page Table Management

• Large page table sizes necessitate hierarchical or hashed page table strategies to reduce overhead.

Swapping

• Allows processes to temporarily be moved from memory to maintain total process space despite limited physical memory.
• The process may not return to the same physical address depending on the address binding method.

Context Switching with Swapping

• High context switch time if a process is not in memory, necessitating a swap with considerable time overhead.