Andrea Garbugli

Building real-time edge platforms @hazeware_it

[Andrea Garbugli @a_garbugli]

6/ What's next? 🚀

Now we know the OS hides hardware states inside `/sys`. But reading raw sysfs files manually is painful.

Tomorrow, we’ll drop into the CLI and look at how to actually parse and explore this hardware tree using `lspci`.

Follow along so you don't miss it! 👇

[Andrea Garbugli @a_garbugli]

Linux Networking Deep Dive: How the OS Sees Hardware 🧵

Let's start from the absolute bottom of the stack. A network packet hits the physical NIC, but how does the Linux kernel actually see and interact with that metal?

It doesn't use magic. It uses high-speed buses and fake filesystems. Let's dig in 👇

[Andrea Garbugli @a_garbugli]

The Journey of a Packet Through the Linux Networking Stack 🧵

Every time you open a browser, stream a video, or ping a server, a packet travels through several layers of the Linux kernel before it ever touches the wire.

Most devs use the network daily without knowing what happens under the hood.

Here is a simple map of that journey, top to bottom:

[Andrea Garbugli @a_garbugli]

5/ The true power of sysfs ⚡

`/sys` isn't just read-only. It's how you manage hardware directly.

You can write to these files to manually bind or unbind drivers from a NIC. This is incredibly useful for troubleshooting, passing a PCIe device directly to a Virtual Machine, or even developing custom drivers!

[Andrea Garbugli @a_garbugli]

Curious how massive `struct sk_buff` actually is? Check out the full source code on Elixir Bootlin. (Btw, if you haven't used Elixir, it's an absolute lifesaver for navigating and cross-referencing Linux kernel source code right in your browser). 🔗https://t.co/YSxIn5OUrs

[Andrea Garbugli @a_garbugli]

The Journey of a Packet Through the Linux Networking Stack 🧵

Every time you open a browser, stream a video, or ping a server, a packet travels through several layers of the Linux kernel before it ever touches the wire.

Most devs use the network daily without knowing what happens under the hood.

Here is a simple map of that journey, top to bottom:

[Andrea Garbugli @a_garbugli]

3/ The Protocol Stack📚

Once data is written to that FD, we enter the Protocol Stack. Your raw data is encapsulated: wrapped in a protocol header (TCP/UDP), then an IP header, and finally an Ethernet frame.

To do this efficiently, Linux uses a core structure: `sk_buff`.

(Check the attached image for a highly simplified look at this massive C struct 👇)

[Andrea Garbugli @a_garbugli]

8/ What's coming next?

This is just the 10,000-foot view. In the coming weeks, I’ll drop deep-dive threads on each layer.

We'll build virtual networks, play with tc, explore Time-Sensitive Networking (TSN), and bypass the stack entirely with DPDK/XDP.

Follow along to see what's under the hood of Linux networking and how to optimize for low-latency! 👇

[Andrea Garbugli @a_garbugli]

7/ The NIC: Not a dumb pipe 🧠

The NIC picks it up autonomously, transmits it, and fires an interrupt.

Modern NICs are capable and complex beasts:
- Compute checksums
- Segment large TCP buffers (TSO/GSO)
- Steer packets to specific CPUs
- Nanosecond hardware timestamping

[Andrea Garbugli @a_garbugli]

@falconerd @japaneseremedy Is it?
- Node, Bun, Deno...
- React, Vue, Angular, Svelte...
- npm, yarn, pnpm (the 2GB node_modules)
- Try CSS without a framework and you literally DIE.

(Just kidding... mostly 😂)

[Dylan Falconer @falconerd]

How to learn game programming:

Old way
- Unity tutorials
- Asset store plugins
- Following YouTube step-by-step
- Never understanding what's happening

New way (ancient way)
- Build from scratch in a systems language
- Implement collision yourself at least once
- Break things, then fix them
- Actually understand your game

Understanding > abstraction

[Andrea Garbugli @a_garbugli]

Really important to understand this. AI is a great tool for learning, but you can’t skip the effort. If you don't put work into what you do, you lose the ability to do it yourself.

[ℏεsam @Hesamation]

this is in fact true. MIT did a complete research on the effect of AI on your cognitive abilities and i’ve never looked at AI the same way since then:
> LLM use accumulate cognitive debt
> the more you rely on AI the worse you get at thinking without it
> you stop exercising cognitive muscles, they grow weaker, you get lazy
the goddamn thing is worse than narcotics if you really think about it. i hope this research is wrong.