Virtual Address Spaces

The illusion of private memory

Each process sees a contiguous virtual address range (typically 0 to 2⁴⁷–1 on 64-bit Linux user space). Code, heap, stack, mmap regions, and shared libraries occupy separate virtual ranges. Process A at virtual address 0x1000 maps to completely different physical frames than process B at 0x1000 — isolation by design.

Typical layout (conceptual)

Region	Grows	Contents
Text (code)	Fixed	Read-only executable instructions
Data/BSS	Fixed	Initialized and zero-initialized globals
Heap	↑ upward	malloc allocations
mmap arenas	Variable	Shared libs, file mappings, large allocs
Stack	↓ downward	Local variables, return addresses

ASLR randomizes base addresses to frustrate exploit chains — /proc/<pid>/maps shows the actual layout.

Why virtual addresses exist

• Isolation — processes cannot read each other's memory without mmap(MAP_SHARED) or ptrace.
• Simpler linking — every binary can assume it loads at the same virtual base.
• Sparse address spaces — reserve 1 TB virtual for a memory-mapped file without committing physical RAM.
• Overcommit — allocate virtual pages now, back with physical frames on first touch.

Senior-level signal

A segfault is the kernel killing your process for an invalid virtual access — unmapped page, write to read-only, stack overflow into guard page. dmesg may show the faulting address; compare against /proc/<pid>/maps to distinguish null-pointer bugs from stack exhaustion.

Where this goes next

Paging & Page Tables explains the data structures that translate those virtual addresses into physical frames — and how page faults make demand paging work.