Index

Source Code Study

Version History

Date	Description
Oct 16, 2021	Move compilers section to a separate file
Dec 7, 2020	add sanitizers section
Sep 13, 2020	some notes for python; add tcpstat
Jul 26, 2020	Add OpenJDK! Hinted by Hacker News :)
Jun 2, 2020	Add librcu
Apr 26, 2020	Add wayland, X, gnome, gtk etc
Apr 10, 2020	add graphics section
Apr 6, 2020	add verbs perftes
Mar 3, 2020	add FreeBSD, some fpga stuff
Feb 4, 2020	add io_uring, firecracker
Jan 31, 2020	Add some good stuff
Jan 18, 2020	Initial

Beautiful code is art. This page documents all the interesting & practical software/hardware/firmware I came across during my work.

• Nutrition
• Operating Systems
• Network
• Virtualization
• Compilers
• Bootloader and Firmware
• Web Servers
• KVS
• Databases
• RDMA and More
• Graphics
• FPGA
• Sanitizers

Nutrition

Projects supporting our day-to-day work.

• GNU glibc: libc, elf, and dynamic linker
  • It is the default C library used by almost everyone
  • It includes ld.so, the dynamic linker
  • I wrote some notes about GOT/PLT and explains what has happend before main() is called.
• GNU binutils: gas, static linker, and more
  • This repo has a lot commands like as, ld, objdump, nm and so on
  • ld is static linker and I like the magic of its linker script
  • I guess another useful repo is elfutils
• C Library
  • GNU glibc used by major Linux distributions
  • musl libc is a small libc impl used by Alpine Linux. Clean code.
  • uClibc is a small libc targeting embedded cases
  • bionic is Android’s C library, math library, and dynamic linker
• [C++ Library]
  • NVIDIA libcu++
• strace
  • System call tracer at userspace
  • I’ve designed one for LegoOS in kernel space
• Unix Commands
  • Of course almost all other listed repos in this section have some sort of commands. But they are not essential. The following repos have the essential UNIX commands like ls, cat. It’s not possible to go through all of them. But rather, I think they serve as references when we want to know how certain things are implemented (e.g., how dmesg get kernel log).
  • BusyBox
  • GNU Coreutils
  • util-linux
  • FreeBSD and its friends
• Tools
  • tmux
  • git
• Editors
  • vim
  • neovim
• C for life
  • Some small and useful C projects
  • cJSON: A lightweight JSON parser in C.
  • userspace-rcu: A userspace RCU implementation library.
• Outliers
  • CRIU: Checkpoint and Restore in Userspace
    • The reason I love this repo is because it has so many interesting pieces on how to interact with kernel, save states, and restore them. In addition, it shows how to properly use many less well known syscalls.
  • GRUB2: bootloader
    • Learn how modern bootloader works.
    • Detailed analysis of Linux booting sequence (how it transit from real-mode to protected mode, and finally to 64-bit mode, how to navigate Linux source code etc.)
  • FFmpeg
    • FFmpeg project is famous for its clean and neat C code.
    • Besides, this project is used by a lot online video service companies
  • io uring
    • user liburing
    • kernel io_uring.c

Operating Systems

See here.

Network

• iperf3 is a TCP, UDP, and SCTP network bandwidth measurement tool
• tcpdump
• iputils (arping, ping, etc)
• scapy: Python-based interactive packet manipulation program & library. Very neat
• tcpstat: C-based simple tool that could dump network traffic. Seems using pcap interface, the one used by tcpdump?
• Also checkout FreeBSD as it has tools like ifconfig, if.
• OpenSSH is our ssh!
• OpenSSL

Virtualization

Also see: http://lastweek.io/notes/source_code/virt/.

• libvirt: virsh and more
• QEMU
• Firecracker
• rust-vmm
• cloud-hypervisor
• Containers
  • runc in go.
  • containerd in go.
  • docker in go.
  • k8s in go.

Compilers

See here.

Bootloader and Firmware

See here.

The open-source firmware landscape:

FPGA

• My own Collection
• My own Paper Readings
• Partial Reconfiguration
  • Partial Reconfiguration Building Framework
  • Intepret Xilinx Bitstream
  • HLS-based ICAP Controller
• Network
  • Corundum: an FPGA-based NIC
    • This is THE BEST network stack out there.
    • This is not simply a network stack, it is a NIC.
    • So what makes a NIC? First, PHY and MAC are basic. Second, PCIe connection between host and board. Third, DMA using PCIe, for TX and RX packets between host and board. Fourth, a host NIC driver; Fifth, some opt modules at NIC.
    • This project has it all. Most amazingly, it works on so many boards.
    • They have an FCCM‘20 paper (finally!) describing the small modules inside.
  • Verilog-Ethernet
    • Self-made PHY, MAC IPs, ARP, IP, UDP stack
    • This is also used by the Corundum project.
  • Limago, HLS-based 100 GbE TCP/IP
  • FPGA Network Stack
    • This one came from ETH as well.
    • This one is used by many papers, as far as i know, StRoM, EuroSys‘20.
    • It’s mostly HLS-based. And has ETH/IP/UDP/TCP, RoCE v2 stack.
• Simulation, Synthesis, and P&R
  • Icarus iverilog. iverilog is a compiler that translates Verilog source code into executable programs for simulation, or other netlist formats for further processing man page.
  • VMware Cascade. Just-in-time compilation for Verilog, what a brilliant idea.
  • Verilog-to-routing.
    • Synthesis (ODIN II)
    • Logic Optimization & Technology Mapping (ABC)
    • Placement and Route (VPR)

Web Servers

• Apache httpd
• nginx

Key Value Stores

Point of interests: 1) in-memory, and can it extend to use disk/ssd? 2) persistence support 3) network support

• RocksDB: A persistent KVS for Flash and RAM Storage. C++
• LevelDB. C++
• Memcached. C
• Redis. C
• etcd: Distributed reliable KVS. Go

Databases

• MySQL
• PostgresSQL
• Yugabyte, distributed SQL

RDMA and More

See here

Graphics

More here

• X Server and Wayland
  • X is being replaced by Wayland now..
  • Wayland code seems clean
• xvnc
  • xvnc and its friends, are sitting on top of display manager (i.e., X/Wayland). They are clients of X/Wayland, but they act as X/Wayland servers for upper layer application such as GTK/Qt.
  • It’s a middleman, bringing network between X and GTK.
  • TigerVNC, TurboVNC and so on.
• GNOME Shell and GTK
  • GTK’s default backend is X.
  • GNOME shell is a layer on top of GTK+. Similar for KDE/Qt.
• xRDP, an RDP server. In C
• FreeRDP, client and server. In C
  • Took a brief read of the code, it’s super neat. Should take a serious look sometime.
• Vulkan/OpenCL
• Proton

The landscape:

Sanitizers

There are many tools in both user and kernel space helping programmers identify various issues early on. Those issues including memory safty issue, threading issue, and others.

Personally I have not used these tools a lot. But I am very interested in them. I think they could greatly improve productivity.

TODO: https://github.com/google/sanitizers