Memory Management

Overview

What is memory management?

1. Physical and virtual memory

2. Allocation

3. Paging

Just scratching the surface!

Virtual memory

We start without any: this is NOMMU mode

Address space isolation

Clone system call

• Demo: man 2 clone

• Aside: See posix_spawn(3)

Question

How is fork defined?

Answer

It's just clone

kernel/fork.c

Namespace interlude

clone(CLONE_VM)

1. CLONE_VM: The new process shares the same memory space as the parent.
2. CLONE_FS: Shares file system information.
3. CLONE_FILES: Shares file descriptors.
4. CLONE_SIGHAND: Shares signal handlers.

Namespaces

Isolated view of system resources

• Mount Namespace (mnt) – Isolates filesystem mount points.
• Process Namespace (pid) – Provides separate process ID trees.
• Network Namespace (net) – Gives processes their own networking stack.
• User Namespace (user) – Provides separate user and group IDs.
• UTS Namespace (uts) – Isolates hostname and domain name.
• IPC Namespace (ipc) – Isolates inter-process communication mechanisms.
• Cgroup Namespace (cgroup) – Provides separate views of control groups.

Demo: unshare

sudo unshare --pid --fork --mount-proc bash

--pid: Creates a new PID namespace.

--fork: Forks a new process to run the specified program (in this case, bash).

--mount-proc: Mounts a new /proc filesystem for the new PID namespace.

Demo: unshare

Look at ps auxf inside and outside the new shell

Find the external PID X of the new internal PID 1

Look at both of sudo ls -l /proc/{$X,$$}/ns

Question

Can we achieve address space isolation without the MMU?

Answer

No

What can a nommu do?

Demo: NOMMU lack of memory isolation

Enabling the MMU:

1. Enable the MMU (disable RISCV_M_MODE, enable MMU in menuconfig)

2. Recompile the kernel, init, and run

3. Successful segfault!

Types of kernel addresses in MMU mode

Logical: fixed offset from physical memory

• This is what phys_to_virt does

Virtual: has entry in page tables

Slab allocators

SLOB

* Simple list of blocks

* Slowest, oldest, simplest

* Uses global `slob_lock`

Slab allocators

SLAB

* Newer, but not the best

* Still locks

* Improved performance via caching and per-cpu lists

Slab allocators

SLUB

* fastest, newest

* Lockless fastpath for alloc/deallocs

* Locks only when crossing CPU boundary

Slab allocators

Demo: /proc/slabinfo

To set allocator: add slab_allocator=sl{u,a,o}b to kernel command line at boot

Collectively, these are the "slab allocators" or "slab layer"

Paging, page reclaim, swap

What weighs more: 10G of physical memory of 10G of virtual memory (joke)

Paging, page reclaim, swap

Technically: virtual memory requires pagetables so there is some overhead

Paging, page reclaim, swap

What's bigger: physical or virtual memory?

Paging, page reclaim, swap

Answer: virtual memory. Why?

Paging, page reclaim, swap

On 64 bit system each process had maximum theoretical address space of size 2^64!

• Most real systems only use 48 bits, some up to 57

• Therefore: we need to swap out pages so physical memory doesn't get full

What is page reclaim

When kernel is under memory pressure:

• Toss File-backed memory pages away

• If pages are dirty, write back to backing file first

• Swap out Anonymous pages

What is swap?

Move memory pages into swap space

• Demo: look at free -mh

Can be a swap file, partition, or compressed RAM

• Demo: see zram with lsblk

Uses least recently used (LRU) algorithm

OOM: Out of Memory

Q: What happens when you run out of memory?

OOM: Out of Memory

A: The out of memory killer is activated to kill a memory-hogging process

Demo: look at mm/oom_kill.c

Page allocation: buddy system

Pages allocation in power-of-two groups

• Use get_free_pages() API

• You can go around the slab allocators

Summary

• fork() is a mirage

Summary

• Virtual memory is not strictly required

Summary

• Virtual memory provides important protections

Summary

• The kernel contains multiple memory allocators

Summary

• The slab allocators rely on the page allocator on the backend

Summary

• There are many aspects to the kernel memory management system beyond this intro

Summary

• We are only scratching the surface

END