It hasn’t been so far away since I was extremely excited for starting to work on that awesome free software project, but now I’m just at the end of that outstanding journey. I ever dreamed in become a free software contributor, and now I can just confirm that it’s what I want to my life.

During EVoC, I could incredibly grow myself into someone who can understand how free software are developed and be confident to start contributing with. Now, I expect that it may be just the start of a long standing journey on the world of free software development, where I aim to keep me always actively working on.

The goals of my EVoC project

Just to recapitulate a bit, the main goal of my EVoC project was to increase the DRM code coverage on the kernel by writing KUnit tests. To do it, I chose select some functions from drm_framebuffer.c to start testing. However, over the project I changed it a bit, since I became to understand a little more about these functions and noticed some better candidates to be tested instead.

Everything was a challenge

Since the start (and even now) I was absolutely concerned about my programming skills and now I’m glad I could come from never programmed any real world software to sending patches to the Linux kernel. This is basically the resume of all that work, I would never imagine that I could start to contributing to such a big and complex piece of software, but the EVoC showed me the opposite.

Over the EVoC, the first thing I learned was that it’s simply impossible to learn everything you think you should know. When starting to work, I faced every video, blogpost and documentation about DRM and the only feeling left was: I know nothing, but now I can complement: I know nothing but an infinitesimal part more then I knew at the start. That little is, in true a huge amount of knowledge, which is what really allowed me to contribute to the kernel. Therefore, what remain is: focus in the small part you wanna work and just start. You’ll learn a lot by doing things, as simple as it may be.

What I have done

At the moment, I was able to contribute with the following patches to the Linux kernel:

1. drm/tests: Fix swapped drm_framebuffer tests parameter names
2. [drm/tests: Stop using deprecated dev_private member on drm_framebuffer tests]
3. [drm/tests: Add parameters to the drm_test_framebuffer_create test]
4. [drm/tests: Add test case for drm_internal_framebuffer_create()]
5. [drm/tests: Add test for drm_framebuffer_check_src_coords()]
6. [drm/tests: Add test for drm_framebuffer_lookup()]
7. [drm/tests: Add test for drm_framebuffer_init()]
8. [drm/tests: Add test for drm_framebuffer_free()]
9. [drm/tests: Add test for drm_mode_addfb2()]
10. [drm/tests: Add test for drm_fb_release()]
11. [drm/tests: Add test for drm_framebuffer_cleanup()]
12. drm: Remove plane hsub/vsub alignment requirement for core helpers
13. drm: Replace drm_framebuffer plane size functions with its equivalents
14. [drm: Add kernel-doc for drm_framebuffer_check_src_coords()]

However, just that first one was accepted and I’m currently working on appling proposed review changes on the patches until they are accepted.

With all the patches I made, the drm_framebuffer.c line coverage came from 18.1% to 44.9%, while the function coverage came from 8% to 52.2%. Despite this file not being entirely covered by unit tests, the outcome coverage is very significant, since the parts left untested are, in majority, legacy code and ioctls, these which in turn is much better tested in integration tests instead, like the ones present IGT.

What was left over

Despite changes from the proposed functions to test to what I really tested, I tried to maintain at least a one-by-one correspondence over them, but there was one function that was left out without any other tested in its place, which is the drm_framebuffer_remove. It’s, of course, partially tested by another test, since it’s split into two different static functions, one for legacy devices and other to atomic capable ones. I spent a good time trying to figure out how to test that atomic part but I was unsuccessful. Of course, I could do like all the other functions I decided to not test and put another in its place, but that in specific would be a good function to test, so I spend a good time trying it instead, but when I figured out that I wasn’t capable of do it, it was too late to try testing another one.

Acknowledgments

I would like to thanks a lot the support I had from my mentors Tony, Maíra and Tales. They played a very important role guiding me through that project and I’m very grateful for all the help I get. I would also like to thanks my colleague Grillo for all his companionship and help.

Moreover, I would like to thank the X.org foundation for accepting and funding the project, without this it would be impossible to me to dedicate myself exclusively to a free software project.