Linux Kernel Monkey Log

Random bits from Greg Kroah-Hartman

Helping Out With LTS Kernel Releases

A recent email thread about “Why isn’t the 5.10 stable kernel listed as supported for 6 years yet!” on the linux-kernel mailing list ended up generating a bunch of direct emails to me asking what could different companies and individuals due to help out. What exactly was I looking for here?

Instead of having to respond to private emails with the same information over and over, I figured it was better to just put it here so that everyone can see what exactly I am expecting with regards to support in order to be able to maintain a kernel for longer than 2 years:

What I need help with

All I request is that people test the -rc releases when I announce them, and let me know if they work or not for their systems/workloads/tests/whatever.

If you look at the -rc announcements today, you will see a number of different people/groups responding with this information. If they want, they can provide a Tested-by: ... line that I will add to the release commit, or not, that’s up to them.

Here and here and here and here and here are all great examples of how people let me know that all is ok with the -rc kernels so that I know it is “safe” to do the release.

I also have a few companies send me private emails that all is good, there’s no requirement to announce this in public if you don’t want to (but it is nice, as kernel development should be done in public.)

Some companies can’t do tests on -rc releases due to their build infrastructures not handling that very well, so they email me after the stable release is out, saying all is good. Worst case, we end up reverting a patch in a released kernel, but it’s better to quickly do that based on testing than to miss it entirely because no one is testing at all.

And that’s it!

But, if you want to do more, I always really appreciate when people email me, or, git commit ids that are needed to be backported to specific stable kernel trees because they found them in their testing/development efforts. You know what problems you hit better than anyone, and once those issues are found and fixed, making sure they get backported is a good thing, so I always want to know that.

Again, if you look on the list, you will see different companies and developers providing backports of things they want backported, or just a list of the git commit ids if the backports apply cleanly.

Does that sound reasonable? I want to make sure that the LTS kernels that you rely on actually work for you without regressions, so testing is key, as is finding any fixes that are needed for them.

It’s not much, but I can’t do it alone :)

So, 6 years or not for 5.10?

The above is what I need in order to be able to support a kernel for 6 years, constant testing by users of the kernels. If we don’t have that, then why even do these releases because that must mean that no one is using them? So email me and let me know.

As of this point in time (February 3, 2021), I do not have enough committments by companies to help out with this effort to be able to say I can do this for 6 years right now (note, no response yet from the company that originally asked this question…) Hopefully that changes soon, and if it does, the release page will be updated with the new date.