CS240lx: advanced operating system lab, accelerated

Engler + Akshay, Spr 2022.

tl;dr index:

• This is a lab class and here are the labs.

• Course website.

• Course newsgroup.

• To get the repository:

  git clone git@github.com:dddrrreee/cs240lx-22spr.git
  

  Put the path to the repo in your shell startup file:

  E.g. for tcsh I put the following in my .tcshrc file (and did source ~/.tcshrc after):

  setenv CS240LX_2022_PATH /home/engler/class/cs240lx-22spr/
  

  If these steps worked, you should be able to type make and have it work.

This is a implementation-heavy, lab-based class that will cover similar topics as CS240 --- threads, virtual memory, file systems and distributed systems --- but by writing code versus discussing papers. After taking an initial operating systems course you are often left in the odd situation of having spent significant time getting the background knowledge needed to do interesting actions, but not being left with enough time to actually do them. This course attacks this problem by assuming basic knowledge and using it as a springboard to advance quickly through a set of powerful, useful techniques that even advanced practioners can be ignorant of.

We will write custom code from scratch for the widely-used, ARM-based raspberry pi; our code will run "bare-metal" without an operating system, which means we can do interesting things without constantly fighting with a lumbering OS that cannot get out of its own way.

By combining research insights (many unpublished) and our our lightweight bare-metal code we will be in the unusual position of having foundational abilities that most people have assumed are in-practice impossible. As one example, we will implement fast, flexible exception handling and then use this ability to build a variety of tools that find race conditions, check code correctness, and memory corruption. By using exceptions rather than binary rewriting we can build tools that find similar errors as Purify / Valgrind, but can be implemented in merely hundreds rather than hundreds of thousands of lines of code; our tools will also be faster and more extensible.

Our initial plan was to do ten projects, one per week, where each project covers two labs of (at a minimum) several hours each and a non-trivial amount of outside work.

The course workload is significant, but I will aim to not waste your time. CS140E is strongly encouraged as a prerequisite, but a sufficiently talented and motivated implementor can make up for its lack.

I'm teaching this course because doing CS140E was was so much fun. Each time I drove into lab, I would get an idea for one or two additional labs I really wanted to do, and each time I wrote up an lab there would be a long NOTES.txt of additional ideas. After last year and this, I have a big document of all the possible ideas. This class will implement the ones that seem the most interesting.

Also, we have Akshay for another quarter as a TA so it seems sensible to exploit that happy fact!