Linux Kernel Explorer
reverser.dev514 points by tanelpoder 17 hours ago
514 points by tanelpoder 17 hours ago
Very neat. It reminds me a bit of how pages in the Talmud are laid out. From a gentile perspective, it was very interesting to me to see how hundreds (thousands?) of years of commentary are contained within the same page.
https://www.reddit.com/r/interestingasfuck/comments/1acgks3/...
https://triberuth.wordpress.com/2016/09/23/my-talmud-layout-...
Code isn't linear the same way, and pages don't make as much sense, but that idea of layers of commentary rings out in this Linux Kernel Explorer as well. I very much like the notes on the side!
Talmud - the original hypertext (tm)
(*maybe, not 100% sure)
see also Pi, the movie, although it's more about numbers and the Kabbalah :) https://en.wikipedia.org/wiki/Pi_(film)
Long-time HN lurker here! Was excited to see this discussion around my major interests of Talmud, Kabbalah, and tech.
There are a lot of misconceptions and mystique surrounding the Talmud. I'd like to take the opportunity to clarify some fundamental aspects, as relates to the discussion here:
The famous "Talmud page" (discussed in the links in the parent comment) was set by a Christian printer in the 16th century.
It emulated a common layout in medieval Christian manuscripts for Christian primary texts and commentaries [0].
The analogy of the Talmud to a hypertext isn't especially apt, IMO. The Talmud indeed extensively cites Bible and Mishnah, and uses lots of technical terms. In this regard, a better analogy is to legal literature (which is what the Talmud in fact is). While being couched as a (fictional) "conversation"/dialogue between rabbis who lived over the course of ~400 years (100 CE to 500 CE).
In fact, Kabbalah (as another commenter mentioned) is a better example of a “hypertext,” since it’s full of recurring symbols that point to different Sefirot and other core concepts.
(By way of credentials: I hold an MA in academic Talmud and Kabbalah, write on these subjects in several venues, and have presented at academic workshops. Over the past two years, I’ve also been developing digital-humanities projects related to this work.)
References:
[0] https://seforimblog.com/2023/06/from-print-to-pixel-digital-...
[1] https://www.ezrabrand.com/p/beyond-the-mystique-correcting-c...
My message here is off-topic, probably a rule violation, but…
I love this. I love how the users of Hacker News provide deep, real insights on pretty much any topic. Thank you!
> The analogy of the Talmud to a hypertext isn't especially apt, IMO.
Isn’t it? Every page of the Talmud includes marginal notes (Masoret HaShas, Ein Mishpat, Torah Or) giving cross-references to relevant parts of the Torah, Talmud and other legal codes. In a web-based version I think it would be natural to represent those with hypertext.
>"Isn’t it? Every page of the Talmud includes marginal notes (Masoret HaShas, Ein Mishpat, Torah Or) giving cross-references to relevant parts of the Torah, Talmud and other legal codes. In a web-based version I think it would be natural to represent those with hypertext."
True, and the website "Al Hatorah" indeed does that, for the marginal notes that you list. See, for example: https://shas.alhatorah.org/Gemara/Berakhot/2a
But my point is that those marginal notes are an artifact of the 16th century print edition. It's not anything inherent in the Talmud text.
The famous 16th-century Mikraot Gedolot edition of the Bible also features extensive marginal notes (the Mesorah) which function much like a dense network of cross-references.
In fact, the Mesorah is a medieval work (drawing on ancient sources) and is arguably was one of the most elaborate systems of cross-referencing found anywhere, at the time it was promulgated.
This differs from the Talmud’s cross-referencing, which doesn't predate the printed edition (as I note in the Seforim Blog article; the page citations are reliant on the universal page numbers that started from the first print edition).
> But my point is that those marginal notes are an artifact of the 16th century print edition. It's not anything inherent in the Talmud text.
OK, fair enough, if ‘the Talmud text’ is taken to be only the Mishna and the Gemara. (Though when I think of the Talmud it’s the printed edition that comes to mind, with all its accompanying commentary.)
EDIT: I had a look at your blog and saw you actually addressed this exact point already: https://www.ezrabrand.com/i/162112983/myth-the-talmud-is-div...
That's awesome!! Thank you very much! I would have next asked you, what do you think of those apps for studying the Talmud, https://www.sefaria.org/app et al, but in those links you already mention it. Looking forward to reading these, thanks!
Definitely!
Sefaria is incredible, it's revolutionized access to classical texts. And their API gives full and complete access. My vibe-coded Talmud reader website fetches Talmud, Bible, and translations from Sefaria, you might be interested in checking it out :)
Source code here:
https://github.com/EzraBrand/replit-chavrutai-2
I've been vibe-coding it over the last few months using Replit, it's been a really cool experience
I must be missing something, we had kernel source browsing online for years, e.g. https://elixir.bootlin.com/linux/v6.17.9/source/fs, this is another one, what extra features it provides. in fact the bootlin one provides some search functions but I did not spot that here. I was expecting some LLM explanation for the code, or dependency graphs etc in the AI era for a new kernel navigator...
You: “Wow, we are surely in the golden age of AI! I can’t wait to see LLM-guided explanations for the code or a new kernel dependency navigator powered by AI…”
The AI Era: “Meh, best I could do is AI rehashes of the same old functionality in a different web framework.”
I love tools like this. I remember using a similar one (made by Red Hat IIRC) that I used to look inside the sources for the Brazilian voting machine (I was asked to map duplicate files and functionality and simplify the codebase) in 2002. It was a desktop app with a very Motif interface.
> API rate limit exceeded for 106.51.68.199. (But here's the good news: Authenticated requests get a higher rate limit. Check out the documentation for more details.)
The GitHub APIs that you are using to list files are getting rate-limited in my case. If somebody else is also facing this issue, just use a VPN or something like Cloudflare Wrap to change your ip - this should fix the issue.
P.S If you are the main dev, giving an option to the visitor to sign in using Github or use a caching layer will be really helpful to make this accessible for the new user.
Hey! Thank you for catching this issue and reminding me of the super naive implementation I went with. I will improve this! :)
It’s often better to overlay caching and other tricks on top of naive implementations than making the implementation more complicated.
* until you get caching layers more complicated than the more complicated implementations :) Imo strong implementations can be much simpler and more predictable than caching.
I love this idea. I like exploring code of interesting projects even if I don't intend to ever work on them, but in complex software I don't know much about it's hard to even find where are the most important basic parts. This allows me to easily find and see how some things in the kernel look like.
Found a bug: in the Chapter 2, when I click on "open" next to "mm/" or other dirs, I get an error: " Failed to load file - Invalid file response from GitHub API - File: mm/". I guess it's cause it tries to open the dir as a file, instead of something like pointing at the dir in the dir tree?
Yes it's good idea and nicely executed.
For the same reason I created a couple of projects with the same goal of lowering the barrier to entry to the linux kernel:
- an app to follow the UDP packet flow in the linux kernel source code: https://dmkskd.github.io/linux-kernel-network-stack-visualiz...
- a (hopefully) simple way to play with the linux kernel source code on a mac: https://github.com/dmkskd/linux-kernel-debugging-on-mac
Kudos to https://github.com/FlorentRevest for all his work in the space
I love your UDP packet flow tool, kudos for making that! I've always wondered how packets move through an OS. Also interesting how many gotos I see all over the place, even though everyone says "goto is the devil". Then again maybe this code was written long before that "proverb" came into existence.
Since you mentioned you love that, I will mention this netfilter packet flowchart by Jan Engelhardt [1]
[1] https://en.wikipedia.org/wiki/File%3ANetfilter-packet-flow.s...
Thanks! I suggest to have a look at a 'live' debugger session video here: https://github.com/dmkskd/linux-kernel-debugging-on-mac?tab=... given your interest
if you have an ARM64 mac (sorry for only supporting this OS at the moment) it should be easy to set this up on your end
You know, I think this Explorer is exactly the tool many of us lacked. Reading the Linux kernel source always felt daunting — thousands of files, confusing paths, complex structure. This feels like a “map” that helps you orient yourself, see how parts interconnect, how VFS works, how modules tie together. Yeah, sometimes a feature breaks (API limits, errors opening directories), but even so — this is a great way to peek “under the hood,” understand the architecture, and take the first step. Big thanks to the folks behind it.
I don't get it. I can't pull master, I can't grep, I can't edit and I certainly can't gmake there.
What's the use?
I believe it's a navigation tool, with pointers to important parts of the code. Useful for those that want to learn about the code base but do not know where to get started.
This looks amazing. Took me some time to find the author (https://fabiomaia.eu). Will drop a message to ask if it'd be possible to have a similar tool for Python. I find it useful to teach new engineers how to find the source code in the cpython repository, check if the code is Python or C, and understand what the code does (some times the docs are a bit lacking/confusing for newcomers).
> Will drop a message to ask if it'd be possible to have a similar tool for Python.
Why not create your own? This is as wild as open-source expectations of a speedy implementation, addressing issues quickly, etc. Folks always want more and more.
Asking a silly question… what piece of kernel code do you find the most awe-inspiring or impressive?
Maybe fs/select.c or the polling machinery.
Agree. The VFS is a delight to read. It's a good intro to the kernel pattern of using function pointers to provide a generic API which other functionality can plug into, simply by implementing the appropriate functions. In this case you'll see all the filesystem drivers implement the VFS operations.
Also: https://elixir.bootlin.com/linux/v6.17.9/source
Elixir works better on mobile despite being around for years.
Thanks for the observation about mobile responsiveness, I will improve it!
Elixir also has working tags instead of just looking at a file at a time, so it is much better at the actual exploring part.
This file from chapter 1 cannot be opened: arch/x86/kernel/entry_64.S ("Failed to load file").
It's also not visible in the directory tree on the left.
(Oddly enough another comment mentioned this already and is voted dead. Why?)
Looks great, looking forward to digging in. I had a similar idea a while back (B.A., before agents) but it seemed too big to take on. Was this able to be realised due to agentic coding?
This is super cool. Love the little icons in the left and would be nice if they were clickable.
I really like this reactive guide style interface, which maybe could be quite a good project idea like mdBook[1] but also you to insert quizzes/examples alongside static notes
How is this different from https://elixir.bootlin.com/linux
Well, they're completely different, but other than that they have similar looking names. I think elixir is more useful for the ways I interact with the Linux kernel. This is attempting to be more of a teaching book. The links to function definitions is what makes elixir useful, and this doesn't have those.
The Elixir cross referencer does not have any special features to help you learn
Elixir is a great tool for exploring and cross-referencing the Linux codebase but for a new person, Linux can come across as very intimidating and complicated. The above explorer makes it very interesting to explore the codebase. This is kind of like an interactive book on Linux internals, with every topic being referenced through code.
Look at the content in the right sidebar, this is like an interactive guide to the kernel.
I don't see all those versions, and for some reasons my fans turned on with the elixir link lol
Love this! Small bug but when changing files, it doesn't reset to position 0 in the file (at least on Firefox on Win11).
Were the quiz questions generated by a human or AI?
In particular, one of the first questions is "What is the fundamental difference between the kernel and a process?" It rejects "The kernel is a special process with elevated privileges" (which is essentially correct) and prefers "The kernel is not a process—it's the system itself that serves processes," which is sort of wrong? The kernel represents itself as a process (process zero), because kernel threads also need scheduling. And it is privileged, obviously.
> The kernel represents itself as a process (process zero)
This isn't true of any modern operating system. Kernel code isn't confined to a single process or even a limited number of processes. Transitioning to kernel mode doesn't necessitate switching to a dedicated process. Prior to the emergence of CPU speculative execution vulnerabilities, it was common for kernel code to be mapped directly into the virtual address spaces of userspace processes.
PID 0 is merely an implementation detail of the scheduler shared among many Unices. It doesn't function like a normal process, nor is it an accurate representation of how the large part of the kernel operates.
Also in Chapter 6 1. What is the relationship between CPU state and kernel state?, it prefers "CPU is stateless; kernel manages state" instead of "They share state equally". I also wouldn't divide it down as "equally" as the kernel manages much more state, but CPUs have registers and cache lines so I wouldn't say they're stateless either.
The very first sentence of the guide is "It's not X—it's Y. It does Z". Safe to assume the entire thing is AI generated.
Can you please elaborate what exactly is the problem with the first sentence?
"The kernel isn't a process—it's the system. It serves user processes, reacts to context, and enforces separation and control."
This is actually based on "The Kernel in The Mind" by Moon Hee Lee. You are welcome to provide feedback.
Is it or is it not AI generated? That's all I said, and you didn't deny it.
Calm down. You had the answer served on a platter.
From "The Kernel in the Mind":
> The Kernel Is Not a Process. It Is the System.
https://www.linkedin.com/pulse/kernel-mind-moon-hee-lee-miwz...
It's X but Y came from elsewhere.
No language server in this explorer, so it is less useful.
Thanks for sharing OP! It seems quite some people liked it, so I'll be listening to feedback and see what to do next. :)
This is very nice. It would be interesting to see the same for other code bases like emacs and vim.
emacs is quite approachable because of its self documenting nature. In practice, if you want to drill down on a particular thing, you need to use some of the describe-* commands. For example want to know how emacs save something?
1) find the command it uses with describe-keybinding, you find the command "save-buffer"
2) "describe-command save-buffer" brings you onto the lisp world where it is defined => "files.el"
3) want to know how a variable is define within files.el? "describe-variable buffer-file-name" and now you are in C territory
4) rinse and repeat with some describe-function when needed
5) get lost onto the beauty of emacs which in my opinion is its interactive / self documenting nature which unfortunatly is not more common in all the software we use
I don't know why you're getting downvoted. It would be wonderful if there was same site for all the big open source projects
when I try to open arch/x86/kernel/entry_64.S. it just says: Failed to load file Failed to fetch file: File: arch/x86/kernel/entry_64.S
maybe its a bug?
Can I somehow deploy it locally?
I appreciate the learning guidance.
What kind of guide says "next topic is scheduler" and just points you to sched.h?
Very interesting! I was kind of expecting that this was going to use LLMs / coding agents to explain the kernel as you stepped through it. But this is a nice app.
I am considering more dynamic "tutors", since this rudimentary version got so much attention! :)
What do you have in mind when you say "step through" the code? Like follow your scrolling of the source code?
Something as simple as an LLM prompt that explains what the file does would go a long way I think.
I love this.
Am I the only one that can't access the website due to certificate problems with the .dev?
I cannot reproduce. Would you mind sharing a bit more? The certificate is handled by Cloudflare Pages.
My bad, the network I was connected to didn't like the certificate for some reason. Cool project!
[dead]
[dead]
