Operating Systems #0 — Getting Started

These are my notes on the operating systems course I took in the first semester of my third year.

The course ran on three tracks: lectures following the slides, weekly labs where you wrote answers to a handful of questions, and three projects where you modified xv6 directly. They weren't three separate subjects. You learned a name in lecture, found that name in the xv6 source in lab, and changed it yourself in the project.

What's in this series

The notes follow those three tracks.

Part	Topic	Source
#1	Computer system overview — registers, fetch-execute, interrupts, cache	Lecture
#2	Operating system overview — multiprogramming, dual mode, system calls	Lecture
#3	Processes and threads — PCB, state transitions, fork/exec	Lecture + lab
#4	CPU scheduling — from FCFS to MLFQ	Lecture
#5	Process synchronization — race conditions, semaphores, deadlock	Lecture + lab
#6	Traps and system calls in xv6 — IDT, alltraps, the boot sequence	Lab
#7	Putting an MLFQ scheduler into xv6	Project 1
#8	Memory management — allocation strategies, page replacement, CoW	Lab
#9	Building LWPs and a process manager in xv6	Project 2
#10	Extending the xv6 file system — multi-indirect, symlinks, buffered I/O	Project 3

xv6

Every lab and project runs on xv6, MIT's teaching Unix — a rewrite of Sixth Edition UNIX for x86. It's under ten thousand lines, so you never spend time hunting for where the scheduler lives. You open proc.c, there's scheduler(), and that really is all of it.

You read in the textbook that "a PCB holds the process state, its registers, and its scheduling information," then open struct proc in proc.h and find those things sitting there as fields. That correspondence was the best part of the course.

How these are written

They aren't transcriptions of the slides. They're closer to a record of what confused me and how I eventually talked myself into an answer. A few paragraphs end on "I still don't fully get this." That's what the notes said at the time, and I've left them in.