Re: Hosting first-party kernel.org container images
Guillaume Tucker <[email protected]> Thu, 19 Mar 2026 11:37:04 +0100
| Newsgroups | dev.linux.lists.kernelci,org.kernel.vger.linux-kernel,org.kernel.vger.workflows |
|---|---|
| Message-ID | <[email protected]> |
Hi Nathan et al,
On 10/03/2026 1:48 am, Nathan Chancellor wrote:
> Hi Guillaume,
>
> On Wed, Feb 25, 2026 at 03:44:13PM +0100, Guillaume Tucker wrote:
>> This boils things down to a few practical options:
>>
>> 1. treating TuxMake images with kernel.org toolchains as the de facto
>> stardard,
>>
>> 2. creating a repository from scratch on git.kernel.org with
>> independent hosting for base images,
>>
>> 3. some middle ground to be defined which would remove the risks
>> associated with third parties without duplicating efforts.
>>
>> I feel it would be good to have more maintainers' feedback though.
>>
>> Nathan, Nicolas, Miguel - what are your thoughts on this?
>
> To be entirely honest, I do not really have a strong opinion here. I
> generally agree with your thoughts around branded container images,
> although I do think that TuxMake's images tend to be fairly lean, so
> those easily could become the recommended images without many downsides
> aside from maybe where they are hosted and how they are maintained.
> Having a clean set of Containerfiles on git.kernel.org does not sound
> like a bad idea, especially if they would be structured in such a way
> that other entities could customize them for their use case further. I
> guess its usefulness really depends on the other stakeholders like
> KernelCI. It seems like there has to be some sort of buy in from the
> kernel.org administrators around hosting built container images
> somewhere on kernel.org though, as I don't think the regular developer
> is going to want to build images locally. Not really sure what that
> looks like.
Thank you for the feedback, it's very valuable even without a strong
opinion regarding the 3 options in my previous email. It also tends
to confirm that the risks I mentioned aren't too critical to get
started, having some developer adoption is more important for now.
With all this in mind, I believe we can follow a step-by-step
approach to start simple and deploy more tailored solutions as and
when the need arises. This can be seen as a bit of a Beta testing
phase until there is enough momentum to look into the longer term.
So here's what I would like to propose for the short term:
1. review current TuxMake kernel.org toolchain images[1] and
determine what needs to be added or changed there (e.g. I don't
see any GCC ones right now)
2. update the scripts/container tool to rely on tuxmake korg images
with default entries in a reference user settings file
3. encourage kernel developers to go and try it out and provide
feedback, maybe even run a small community survey
4. review the situation after a while e.g. once v7.1-rc2 is out and
decide what steps to take next based on community priorities
I believe there aren't any blockers here so we can probably just give
it a try as an experiment and see how it goes, unless someone has
anything else to suggest.
Things like moving the Containerfiles to git.kernel.org or hosting
the images in a dedicated registry etc. should eventually emerge by
themselves if there is a real need for them.
Best wishes,
Guillaume
[1] https://hub.docker.com/u/tuxmake?search=korg