Linux Kernel Mentorship Journey

• Linux

A Brief Intro

Before starting the Linux Kernel Mentorship Program from Linux Foundation , I used to work as a web developer. With how chaotic web development has become , I decided to take a career break. I have been interested in the low-level programming, so I use my time during this career break to learn more about low-level programming.

With that in mind, this blog post’s perspective is from someone who is completely new to kernel development and also new to C programming language. Like many web developer out there, I didn’t use C programming language at work so I am not really familiar with C programming language. But before the mentorship begin, I made a barely working lisp interpreter to make myself familiar with C programming language and linux kernel coding style (using if without curly bracket is a new experience for me).

The mentorship session that I joined in was Linux kernel Spring Unpaid 2026 .

Prerequisite Tasks

There are some prerequisite tasks before we got accepted into the mentorship program. And now you might be wondering, “how did you manage to get accepted even though you are a newbie?”. Well, in case you haven’t noticed, kernel is also a software and if you have developed software before, you can use some of your previous software development experience in kernel development too. There are some differences when developing a kernel and web application for sure, like the testing mechanism. For example, I can’t use the testing framework for web application to test the kernel so I need to figure out how to test the kernel. But the basic stuff like “what should we consider when testing a software?” is the same. I am not sure if this is true, but I feel like the prerequisite tasks for this mentorship testing our intuition as software developer rather than specific to kernel development, and that’s why I can pass the prerequisite tasks.

I can’t disclose the prerequisite tasks, but I suggest you to take a look at Linux Foundation’s A Beginner’s Guide to Linux Kernel Development course . That course is helpful to complete the prerequisite tasks and during the mentorship in general, especially the part to send the patch. I mean, even after you are doing all of those testing and stuff, and then you found out that your patch can not be applied by the maintainer, that would be sad, right? So make sure that your patch can be applied to the linux kernel git tree.

Other than the Linux Foundation’s course, you might also want to take a look at the linux kernel testing guide to get an overview of how linux kernel testing is done.

First Month

After we got accepted into the mentorship program, we have a weekly meeting with the mentors. During those weekly meeting, we can ask the mentors about the linux kernel development process, like how to handle Reviewed-by tag for example.

The most shocking thing about this mentorship is that, we only have one task for the entire mentorship program and that is having a minimum of 5 patches accepted into the linux kernel. There’s no weekly assignments or something like that. As long as you have 5 or more patches accepted during this mentorship program and finish your final report, you can graduate this mentorship program.

Now, here comes the big problem. As someone who is completely new to kernel development, I have no idea what I should do. I don’t even know which subsystem I want to work on. In fact, by the time of writing this blog post, I just learned the concept of virtual memory from book called Operating System: Three Easy Pieces . Yeah, I am that new to this operating system stuff, and the lack of task assignments make this mentorship even more challenging.

So, what did I do? On the first month, instead of looking at the linux kernel code, I take a look at the linux kernel mailing list, especially linux-kernel-mentees mailing list . I want to learn from the previous mentees' patches, what kind of mistake they made and what kind of patch that would be accepted. I am also looking around for resources about contributing to linux kernel as a newcomer. At first, I found a blog post about fixing bug found by syzbot , but at that time I can’t even interpret the stack trace from the crash report (I don’t even know about cpu register until recently, don’t judge me) and that’s why I decided that syzbot’s report is not for me yet. So I keep looking until I found Gustavo’s blog post about fixing coverity issues , and that’s how I made my first contribution.

Lesson from First Patch

Before submitting any patches, I suggest you to take a look at submitting patches documentation on Documentation/process/submitting-patches.rst in the linux kernel git tree.

This is the lesson I got from sending my first patch:

• There is a specific pattern for the commit subject (the one that appear when we do git log --oneline). For example:

staging: rtl8723bs: remove unused offset in phase 2 _BlockWrite()

Notice the staging and rtl8723bs from the example above? The staging there represent the subsystem, and the rtl8723bs represent the device driver. This pattern varies for each subsystems, so you need to check that first before committing the changes.

• When we mention a specific commit, we don’t need to use the full length of those commit hash. We only need 12 characters of the commit hash (this might change as the commit in linux kernel git tree increase). After those commit hash, we need to include oneline summary (or commit subject). For example:

c1314fe4d28f ("staging: rtl8723bs: remove all RT_TRACE logs in hal/ and os_dep/")

• When dealing with private url (like the url that need an account to view the content), we need to put the information from those private url instead of the url itself. For example, because coverity scan need an account to access the information, instead of putting the coverity scan’s report url, we can put the information from those coverity scan’s report like “This is reported by Coverity Scan as Unused Value”. The “unused value” is the information from the coverity scan’s report.

• Don’t forget about the Signed-off-by: tag too, and use your legal name with it. For example:

Signed-off-by: Robertus Diawan Chris <robertusdchris@gmail.com>

This is the first patch that I sent and accepted:
https://lore.kernel.org/linux-kernel-mentees/20260420044651.164450-1-robertusdchris@gmail.com/t/#u

From that first patch, I also learned that when someone replied with Reviewed-by: tag without any feedback, we don’t need to do anything and just wait for the maintainer to respond.

The Next Lessons

As I sent multiple patches (some of them getting a response, some of them lost in the void), I will only pick the patch where I learned the most from.

Let’s take a look at this patch:
https://lore.kernel.org/linux-kernel-mentees/20260513091031.145826-1-robertusdchris@gmail.com/t/#u

In that patch, I give extra information after the triple dash (---) that I don’t have the device and not sure how to test this changes. If we didn’t put that information, the maintainer will assumed that we already test the changes. Where we put the testing information depends on the maintainer’s preferences. Some maintainer prefer to put the testing information in the commit message but other maintainers prefer to put the testing information after the triple dash (---), so we need to adjust the patch to each maintainer’s preferences.

Anything after the triple dash (---) will be ignored when we apply the patch, so we can treat that as a way to communicate with the maintainer outside the commit message, like asking for confirmation or something like that.

The ideal scenario would be to test the changes in a real device, but sometimes not having the device might not be a blocker (unless explicitly said so, like in drivers/staging/fbtft/README). The point is to be open whether we can test the changes or not, and let the maintainer decide whether they want to accept our patch or not.

Other than that, I also add the Fixes: tag. We can put the Fixes: tag when we fix some problem, and in this case the undefined behavior which means anything can happen at that point. The content of the Fixes: tag is the commit which introduces the problem. In this case, the undefined behavior problem appear since commit d6e290837e50.

From that patch, I got a feedback from Amirreza Zarrabi (the maintainer) and after I address those feedback, I can add Reviewed-by: tag in my patch. In this case, I changed my patch to follow the feedback and put the Reviewed-by: tag on the next version of the patch (which is version 2).

Here’s another important lesson, when sending the next version of our patch, we need to send that patch as the new email, not as a reply to the previous version. And inside those new email, we need to put what changed from the previous patch. One of the reason doing this is reducing confusion. As the patch revision grow, if we put the new patch as the reply to the previous patch, at some point someone is gonna be replying to the wrong version. Keep in mind that the maintainers getting hundreds or even thousands of email per day, so if we send the new version of the patch as the reply, our new version of the patch might be buried in the maintainer’s email list. And some tools that the maintainer used, like b4 , might break if we put next version of the patch as the reply.

Make sure to at least wait a few days before sending the new version of the patch. There’s no specific time when we should send the new version of the patch, but my rule of thumb is to wait for at least a day before sending the new version of the patch. It depends on how active the discussion is, if there’s an active discussion, we might need to wait for a week or more.

Here’s the next version of the previous patch:
https://lore.kernel.org/linux-kernel-mentees/20260519020528.133623-1-robertusdchris@gmail.com/t/#u

Here’s another person’s patch with more revision’s version as an example:
https://lore.kernel.org/all/20260609-enable-ice-clock-scaling-v11-0-1cebc8b3275b@oss.qualcomm.com/

Patchwork

Another lesson that I learned is that some maintainers using patchwork to organize the patches they want to handle. So if the subsystem that you worked on have the patchwork setup, like linux kernel build , you can check if your patch in there or not. If your patch is not in the patchwork, it’s likely either the maintainer is not interested in your patch or your submission trigger the patchwork’s filter setup by the maintainer (and the maintainer missed your patch).

Checking the patchwork might resolved some of the anxiety from the newcomer, as most of the newcomer expect instant reply from the maintainer, which might not be the case. The maintainer usually respond to the high priority patches first and respond to the low priority one after the high priority patches has been resolved. So, be patient.

Keep in mind that some of the subsystems did not use the patchwork, so you might not be able to check whether the maintainer is interested in your patch or not.

Summary

What I get from this mentorship is mostly about the process of submitting patches rather than the technical aspect of the kernel development, and that’s what I expected. Submitting patches for the linux kernel can be confusing, so this mentorship helped me understand those process better.

Keep in mind that what matters the most when submitting the patch is expressing our intention. Assume that the maintainer can not read our mind, so add any information that might help the maintainer decide whether to accept our patch or not.

At the time of writing this blog post, I am learning about kdump by using one of the solved syzbot’s report as the test case. The issue has been solved, so what I do is tracking how do we get to the solution using kdump.

I made a simple shell script to create a debian image with kdump setup to reproduce the syzbot’s report. So everytime the kernel panic, we can copy the dump files into our machine and analyze that using crash utility . The script is in here:
https://gist.github.com/bruhtus/00c2e8567f52e1160b9f1bc111a2959c