Posts from 2021
Three years ago, I wrote about a way to purge only changed static files when deploying a static site. It is very useful and I still use it for this website to this day. Its main advantage is that it only needs to be run on deploys. However, its main disadvantage is that it must be run on every deployment. Sometimes, this is not feasible.
For example, I run a bunch of APT repositories on apt.quantum2.xyz. These repositories are constantly being updated by Jenkins and me personally, and using
purge-staticwould require adding a
purge-staticcommand to every script that updates the repositories, which is clearly infeasible. Wouldn’t it be nice to just have a background daemon that purged the CDN cache automatically?
As it turns out, I already wrote it back in 2015 before starting this blog. It was massively out-of-date (until very recently) and required you to use your all-powerful Cloudflare API key, providing a massive attack surface. However, I recently updated it, and hopefully, it will prove useful for you.
Here’s a quick introduction to using it:
Recently, I bought a USB 3.0 2.5 Gbps Ethernet dongle for my Atomic Pi router. This dongle requires a version of the
r8152kernel driver with support for the RTL8156 chipset, which is only added in Linux 5.13. Now, I am running the Debian stable kernel and have no wish to backport the latest 5.13 kernel simply for that one driver. So of course, I came up with an approach to backport a driver from a newer version.
In this blog post, I will walk you through the process of backporting a single kernel module, using the
r8152kernel driver as an example.
It’s not often in 2021 that you find yourself building new kernels, but nevertheless, the occasion comes that you need to either enable a flag—or even worse—patch the kernel. This happened recently: on DMOJ, we recently run into a kernel issue that misreports the memory usage for processes as an “optimization.” For more information about this issue, see the excellent blog post by my friend Tudor. As a result of this, I was forced to build a patched kernel to work around this issue. Since the process was far from easy, I decided to write this blog post to help others in the future.
Building a kernel is not too difficult, actually. The real challenge comes in the form of building the kernel in a maintainable way, which basically means that we should at least build the kernel into an easily installable package. For example, on DMOJ, we manage multiple judge virtual machines, and they all need to receive the same kernel. Furthermore, we want our custom build of the kernel to be distinct from the standard kernels that the operating system offers, as we don’t want a system upgrade to undo the patch that we applied.
In this article, we will explore the process I used to build a custom kernel package on Debian for the scenario described above. This will involve both patching the kernel and subsequently changing a configuration option. Specifically, we will be applying this patch. These instructions should work with minor adaptations for other Debian-based distributions.
Around four years ago, I wrote a blog post about creating vanity
.oniondomains for Tor. To recap,
.oniondomains are special domains understood by Tor, and when visiting these sites, traffic never leaves the Tor network and is end-to-end encrypted, just as if you were to use HTTPS. Furthermore, the server’s identity is completely protected, if you are into that sort of thing. Also, the domain name is linked to the server’s public key fingerprint, and so knowing the
.oniondomain is sufficient to authenticate the server and preventing any MITM attacks or certificate authority compromises that could happen with HTTPS.
I am sure that if you managed a Linux system for a while, you probably have dealt with Unix sockets—special files that act like sockets. You probably also run into permission issues when dealing with these socket files.
In this post, I’ll describe some methods of dealing with these permission issues, and a situation in which each might apply.