think os

Following a trail from a wonderful Julia Evans post led me to Allen Downey’s nice textbook manifesto. Also led me to the nice Think OS book, which seems like a super nice introduction to Operating System principles.

It is short enough (~100 pages) that I wanted to read through it. I’ve spent a good chunk of time absorbing how Operating Systems work by dint of diving into things and working through them, but it would be nice to get a refresher on the basics. There are clearly basic things I do not understand, and this seemed like a good way to explore.

This post is just a running series of notes from me reading it on a nice saturday morning.

Stack vs Heap

This is something that has always bugged me. I’ve understood just enough of this by being burnt with pointers when writing C (and primitive types in the CLR, etc), but was lacking a deep understanding of wtf was going on. The fact that these are just process program segments (like text or data) was quite a revelation :D This stackoverflow answer was also quite nice.

One interesting thing for me to investigate later from the book is how this program:

#include <stdio.h>
#include <stdlib.h>

int global;

int main() {
  int local = 5;
  void *p = malloc(128);

  printf("Address of main is %p\n", main);
  printf("Address of local is %p\n", &local);
  printf("Address of global is %p\n", &global);
  printf("Address of p is %p\n", p);


produces the following output for the author:

Address of main   is 0x      40057c
Address of local  is 0x7fffd26139c4
Address of global is 0x      60104c
Address of p      is 0x     1c3b010

but for me,

Address of main   is 0x5598fc64c740
Address of local  is 0x7ffeacfaf75c
Address of global is 0x5598fc84d014
Address of p      is 0x5598fc85b010

The point of the program was to demonstrate that text (main), static (global) and heap (p) are near beginning of memory and stack (local) is towards the end. While on my laptop it does seem to be the case too, the ‘start’ seems to be much farther out than on the author’s computer. Need to understand why this is the case. I’ve vaguely heard of address randomization & other security measures in OS kernels - maybe related? For another day!

Bit twiddling

I continue to find it hard to care about bit twiddling. Most things do use it of course, but it seems to be abstracted away pretty well without leaking too much (except for things that have their own nuances, like floating point representations).


Nice link to a paper about a common malloc implementation. I also know there are other malloc implementations that programs use (such as jemalloc). Something for me to dive into when I’ve more time.


I didn’t have time to finish it all, unfortunately. But shall come back to it whenever I can!

learning selinux and apparmor

I am trying to understand SELinux and AppArmor, and collecting resources here as I learn. k

SELinux for mere mortals (2014)

This was the first video I watched, and it helped me understanding what SELinux does at a fundamental basic level. It’s probably useless in a container-filled world (where I doubt Fedora shipes pre-configured SELinux rules for my containers), but it helped me think I understood types / labels, so that seems like a positive step?

The fact the presenter keeps saying things like ‘you being a good sysadmin, ssh into the server and edit the apache config file’ is freakin me out. If I’m constantly editing config files on servers manually that seems like a massive failure to me :D How times change!

Docker and SELinux (2014)

This one made a lot more sense to me as an answer to the following questions:

  1. Aren’t containers secure enough? (Partial answer)
  2. What does SELinux do for container security?

It’s convinced me that container -> host isolation and container <-> container isolation provided by SELinux is pretty simple and super useful, and should be turned on.

This talk also showed me this most wonderful coloring book that tries to explain SELinux. If this is all that is to SELinux, it seems pretty simple and useful (for the container use case).

Also, it looks like there are more recent versions of both these two talks - I should look ‘em up!

Securing Linux Applications with AppArmor (2007?)

This is me trying to understand AppArmor, which seems to have lower base of support (just Ubuntu? Maybe SUSE, but idk anyone who uses SUSE) but theoretically simpler (mostly file path based). The video seems to be shot with a potato, so the slides aren’t super clear - but the content is good enough to give me a super general overview.

The biggest thing against SELinux it talks about seems to be ‘SELinux is complex’, and not much else. I don’t know how much I buy that - but then again, I haven’t actually used SELinux anywhere :D

Unlike SELinux, I can actually see AppArmor rules on my local machine (since it is running Ubuntu). Seems fairly readable!

things to build

This is a running list of things I want to build!

There’s an analogous running list of things I want to learn. Things move between them :) I also have higher standards of documentation (other people should be able to use it) before marking these as complete.

  • kubernetes-login A helper to openssh that allows users to log in to a configurable user pod running on a kubernetes cluster. Should ideally support scp / sftp too. Helps get rid of SPOF login nodes

  • just-enough-containment A purely for-learning docker-ish container project written purely in python. Written for pedagogy and personal understanding rather than production use.

python gil resources

I was in a conversation about the Python GIL with friends a few days ago, and realized that my understanding of the specifics of the GIL problem were super hand-wavy & unstructured. So I spent some time collecting resources to learn more, and now have a better understanding!

Python’s Infamous GIL (Larry Hastings)

This was a great introduction to the history of the GIL, why it was necessary & reasons why getting rid of it is complicated.

Understanding the Python GIL (David Beazley)

This has wonderful visualizations that really helped me understand exactly why multi-threaded python behaves the way it does. Multithreading decreases performance, adding more cores decreases performance & disabling cores increases performance :) All of this made vague hand-wavy sense to me before, and make much more concrete sense now.

It isn’t easy to remove the GIL (Guido van Rossum)

A blog post from the BDFL of python, after yet another request to ‘just get rid of the GIL’.

It set the (pretty high) bar for inclusion of a GIL removal patch (that he makes clear he will not write) in Python:

I’d welcome a set of patches into Py3k only if the performance for a single-threaded program (and for a multi-threaded but I/O-bound program) does not decrease.

Not been met yet!

An Inside Look at the GIL Removal Patch of Lore (Dave Beazley)

There was an attempt in about 1999 to remove the GIL - the ‘freethreading’ patch. This is a wonderful analysis of that patch - what it tried to do, why it disappeared, what the performance costs of it were, etc. Something that really stood out to me and makes me feel not very hopeful about GIL removal in CPython was:

Despite removing the GIL, I was unable to produce any performance experiment that showed a noticeable improvement on multiple cores. Really, the only benefit (ignoring the horrible performance) seen in pure Python code, was having preemptible instructions.

This seems to be still true, even in the Gilectomy branch.

Gilectomy (Larry Hastings)

This is the only talk about a recent (~2016) GIL removal attempt.

It is amazing work, but doesn’t give me much hope. There’s been no new commits to the public git repo for about 5 months now, so am unsure what the state of it now is.

There’s probably many more - let me know if you know any, and I’ll update this when I find out more!

Gilectomy - 2017 (Larry Hastings)

PyCon 2017 just happened, and Larry Hastings gave another talk!

It seems to have had a lot of intense work done on it, and the wall clock time graph in it warms my heart! I’ve a little more hope now than I did after the 2016 talk :D

systemd simple containment for GUI applications & shells

I earlier had a vaguely working setup for making sure browsers, shells and other applications don’t eat all RAM / CPU on my machine with systemd + sudo + shell scripts.

It was a hacky solution, and also had complications when used to launch shells. It wasn’t passing in all the environment varialbes it should, causing interesting-to-debug issues. sudo rules were complex, and hard to do securely.

I had also been looking for an excuse to learn more Golang, so I ended up writing systemd-simple-containment or ssc.

It’s a simple golang application that produces a binary that can be setuid to root, and thus get around all our sudo complexity, at the price of having to be very, very careful about the code. Fortunately, it’s short enough (~100 lines) and systemd-run helps it keep the following invariants:

  1. It will never spawn any executable as any user other than the ‘real’ uid / gid of the user calling the binary.
  2. It doesn’t allow arbitrary systemd properties to be set, ensuring a more limited attack surface.

However, this is the first time I’m playing with setuid and with Go, so I probably fucked something up. I feel ok enough about my understanding of real and effective uids for now to use it myself, but not to recommend it to other people. Hopefully I’ll be confident enough say that soon :)

By using a real programming language, I also easily get commandline flags for sharing tty or not (so I can use the same program for launching GUI & interactive terminal applications), pass all environment variables through (which can’t be just standard child inheritence, since systemd-run doesn’t work that way) & the ability to setuid (you can’t do that easily to a script).

I was sure I’d hate writing go because of the constant if err != nil checks, but it hasn’t bothered me that much. I would like to write more Go, to get a better feel for it. This code is too short to like a language, but I definitely hate it less :)

Anyway, now I can launch GUI applications with ssc -tty=false -isolation=strict firefox and it does the right thing. I currently have available -isolation=strict and -isolation=relaxed, the former performing stronger sandboxing (NoNewPrivileges, PrivateTmp) than the latter (just MemoryMax). i’ll slowly add more protections here, but just keep two modes (ideally).

My Gnome Terminal shell command is now ssc -isolation=relaxed /bin/bash -i and it works great :)

I am pretty happy with ssc as it exists now. Only thing I now want to do is to be able to use it from the GNOME launcher (I am using GNOME3 with gnome-shell). Apparently shortcuts are no longer cool and hence pretty hard to create in modern desktop environments :| I shall keep digging!

systemd gui applications

Update: There’s a follow-up post with a simpler solution now.

Ever since I read Jessie Frazelle’s amazing setup (1, 2, 3) for running GUI applications in docker containers, I’ve wanted to do something similar. However, I want to install things on my computer - not in docker images. So what I wanted was just isolation (no more Chrome / Firefox freezing my laptop), not images. I’m also not as awesome (or knowledgeable!) as Jess, so will have to naturally settle for less…

So I am doing it in systemd!

Before proceeding, I want to warn y’all that I don’t entirely know what I am doing. Don’t take any of this as security advice, since I don’t entirely understand X’s security model. Works fine for me though!

GUI applications

I started out using a simple systemd templated service to launch GUI applications, but soon realized that systemd-run is probably the better way. So I’ve a simple script, /usr/local/bin/safeapp:

exec sudo systemd-run  \
    -p CPUQuota=100% \
    -p MemoryMax=70% \
    -p WorkingDirectory=$(pwd) \
    -p PrivateTmp=yes \
    -p NoNewPrivileges=yes \
    --setenv DISPLAY=${DISPLAY} \
    --uid ${USER} \
    --gid ${USER} \
    --quiet \

I can run safeapp /opt/firefox/firefox now and it’ll start firefox inside a nice systemd unit with a 70% Memory usage cap and CPU usage of at most 1 CPU. There’s also other minimal security stuff applied - NoNewPrivileges being the most important one. I want to get ProtectSystem + ReadWriteDirectories going too, but there seems to be a bug in systemd-run that doesn’t let it parse ProtectSystem properly…

Also, there’s an annoying bug in systemd v231 (which is what my current system has) - you can’t set CPUQuotas over 100% (aka > 1 CPU core). This is annoying if you want to give each application 3 of your 4 cores (which is what I want). Next version of Ubuntu has v232, so my GUI applications will just have to do with an aggregate of 1 full core until then.

The two environment variables seem to be all that’s necessary for X applications to work.

And yes, this might ask you for your password. I’ll clean this up into a nice non-bash script hopefully soon, and make all of these better.

Anyway, it works! I can now open sketchy websites with scroll hijacking without fear it’ll kill my machine!


I wanted each tab in my terminal to be its own systemd service, so they all get equitable amount of CPU time & can’t crash machine by themselves with OOM.

So I’ve this script as /usr/local/bin/safeshell

exec sudo systemd-run \
    -p CPUQuota=100% \
    -p MemoryMax=70% \
    -p WorkingDirectory=$(pwd) \
    --uid yuvipanda \
    --gid yuvipanda \
    --quiet \
    --tty \
    /bin/bash -i

The --tty is magic here, and does the right things wrt passing the tty that GNOME terminal is passing in all the way to the shell. Now, my login command (set under profile preferences > command in gnome-terminal) is sudo /usr/local/bin/safeshell. In addition, I add the following line to /etc/sudoers:

%sudo ALL = (root) NOPASSWD:SETENV: /usr/local/bin/safeshell

This + just specifying the username directly in safeshell are both hacks that make me cringe a little. I need to either fully understand how sudo’s -E works, or use this as an opportunity to learn more Go and make a setuid binary.

To do

[ ] Generalize this to not need hacks (either with better sudo usage or a setuid binary) [ ] Investigate adding more security related options. [ ] Make these work with desktop / dock icons.

I’d normally have just never written this post, on account of ‘oh no, it is imperfect’ or something like that. However, that also seems to have come in the way of ability to find joy in learning simple things :D So I shall follow b0rk’s lead in spending time learning for fun again :)

things to learn

Keeping a running list of things I want to learn!

There’s also a list of things I want to build.

  • How to use org mode properly? Should I use it for notes over markdown?
  • Develop a deep understanding of how networks work.
  • How do linux network namespaces work?
  • How to run GUI apps with systemd?
  • What exactly is a ‘tty’?
  • How does HTTP2 actually work?
  • How do X509 / TLS certificates work?
  • How to use cgroups directly?
  • Can I just use emacs terminals for all my terminal needs?
  • How does NFS work, and why is it so crappy?
  • How does ssh work?
  • How does mosh work?
  • How do contact lenses work? HOW DO LENSES WORK?
  • Can I simply run a local DNS recursor on my laptop for performance & blocking me from visiting the orange website?
  • What is SELinux? Why and how would I use it?
  • What is AppArmor? Why and how would I use it, over SELinux?
  • What is seccomp, and when/why/how would I use it?

If you know of resources that’ll help me learn these things, do let me know!

moving to hugo

I’m attempting to now blog at, using hugo rather than wordpress.

Over the last few years, IRC, Twitter & WhatsApp have ruined my public writing. I shall now slowly attempt to bring that back :)

liberal software

I ran into this thought provoking though when randomly attempting to relax this weekend. There’s a summary at LWN if you do not want to watch the talk - but as the lwn summarizer admits, the video definitely conveys things that are hard to capture on text.

The core takeaway for me is to think about:

what is the future of free and open-source software? The answer was: it has no future.

This seems somehow connected to ‘democratizing programming’, which I had earlier given a talk about. Somehow, it feels like there needs to be an update / rebirth of the GNU Freedoms for the world we live in.

my first protest

I went to the protests at SFO last weekend. It was the first real set of protests I’ve been to. I write this to attempt to capture a sliver of what I felt that day.

I was there for about 10h on day 1, and came home exhausted. I went back on Day 2, and this time stayed for much shorter period of time (~4h?) before heading back home.

To everyone who was at the protest even if it does not directly affect you yet - thank you!

To the wonderful amazing women of color who were leading the protest - thank you. You are an inspiration.

To the brass band and the troupe of drummers who showed up - thank you!

To the people I already know from other contexts who I ran into at the protest - thank you!

To those who were protesting for the first time in their lives (as I was) - thank you!

To the fucking ACLU - thank you! I’ll send you all the money I can :)

To the people who provided infrastructure (food, water, snacks, first aid, printers, internet, etc) - thank you!

To the journalists who covered this - thank you.

I woke up that saturday feeling very depressed, angry, and helpless. By sunday night, I was only depressed and angry - but not helpless. The number and variety of people at the protests was very heartwarming, and made me feel distinctly not alone. Before going I was not sure what going to a protest accomplishes. I still do not know - but I know it gave me hope and restored my sense of agency.

I promise I’ll continue doing all I can, even and especially when it is about things that do not affect me yet. I hope you do too.

What hurts the victim most is not the cruelty of the oppressor but the silence of the bystander

- from a sign at the protest