xv6 Assignment 3 — Shared Memory & Multi-Process Logging

📌 Overview

This assignment extends xv6 with shared memory support and builds a multi-process logging system on top of it.
It integrates virtual memory manipulation, page-table operations, and synchronization using atomic primitives.

---

🧠 1. Shared Memory Mechanism

🎯 Objective

Enable one process to map a region of its virtual memory into another process’s address space — similar to real OS IPC mechanisms.

---

🔧 Kernel Additions

map_shared_pages(src_proc, dst_proc, src_va, size)

• Translates virtual→physical pages in the source.
• Maps them into the destination process’s page table.
• Handles alignment (page rounding) and offset correction.
• Copies permission flags and adds a new shared-page marker (PTE_S).
• Updates sz for the destination process.

unmap_shared_pages(p, va, size)

• Safely removes mappings without freeing shared physical pages unless owned by the process.
• Prevents double-frees during process termination.

🛠️ System Calls

sys_map_shared_pages();
sys_unmap_shared_pages();

---

🧪 Test Program: shmem_test.c

Validates:

• Parent/child shared communication
• Correct string exchange
• Proper unmapping and allocation after unmap
• Cleanup behavior when child terminates without unmapping

---

📝 2. Multi-Process Logging System

🎯 Objective

Use the shared memory abstraction to create a concurrent logging buffer written to by multiple children.

---

🔧 Buffer Layout

Each log entry has:

[ 32-bit header | message bytes ... ]

Header format:

• Upper 16 bits → child index
• Lower 16 bits → message length

---

⚙️ Synchronization

Used:

__sync_val_compare_and_swap()

for atomic header claims.

Alignment rules:

addr = (addr + 3) & ~3;

Ensures 4-byte alignment for atomic header writes.

---

🧪 Test Program: log_test.c

• Forks multiple children.
• Parent allocates and maps a shared buffer.
• Children write messages until buffer is full.
• Parent scans and prints messages.

---

🛠️ Skills Demonstrated

• Kernel virtual memory manipulation
• Page-table design and multi-process address space management
• Atomic synchronization in shared memory
• Concurrent system design in a minimal OS
• Debugging low-level interactions between processes

---

📜 License

For educational and demonstration purposes only.