I was having some problems with the latest kernel in clear with my AMD GPU, which needed a newer kernel. I figured I would wait until the newest 5.1 kernel got into clear, but then I found the
kernel-mainline bundle, which let me bump up to the newest kernel. (Check out clr-boot-manager -h to see how to list and switch to new kernels)
swupd search kernel only returns this:
Bundle with the best search result: kernel-pk - Run the Intel "PK" kernel and enterprise-style kernel with backports. This bundle can be installed with: swupd bundle-add kernel-pk Alternative bundle options are kernel-lts-dkms - Support module for building/loading via Dynamic Kernel Module System (DKMS) in LTS kernel. kernel-lts - Run the latest long-term support native kernel. kernel-hyperv-mini - Run the Microsoft Hyper-V mini-os specific kernel.
But from looking in github I see lots o’ kernel bundles.
So, I get the idea behind having cloud specific kernels, and same for virtualization platforms.
kernel-mainline got me a new enough kernel to fix my hardware problem, and I suspect that
kernel-mainline-vanilla is as aggressive but without clear patches.
From watching the videos from this past meetup I think I understand how the QA process works, but I was hoping someone could clarify the thought process behind publishing kernels. I suspect with -vanilla that if it builds it ships and I get to keep both pieces if it’s broken, and probably with mainline too? Also, the lts kernel packages have years attached to them, which don’t line up with what kernel.org, so if someone could explain how the kernel “channels” work that’d be awesome!
Also, how does the publishing of the mainline and mainline-vanilla kernels work? Is it human gated or does a CI/CD bit just monitor the kernel repo and build and push?