Signal

Linux 5.10 and 5.15 reach end of life in December. Here's which of your boxes that actually affects

AusGeek · · 6 min read

Two of the longterm Linux kernels, 5.10 and 5.15, hit their projected end of life on kernel.org next month. If you run a normal distro and install its updates, this mostly isn't your problem, and I'd ignore the headlines telling you otherwise. The machines worth five minutes of your time are the other ones: the SBC running an image you flashed in 2022, the box with a vendor kernel, the appliance nobody has logged into since it was set up.

So here's what's actually ending, who it touches, and how to tell which machines in your cupboard are which.

What's ending

The kernel.org releases page lists the active longterm kernels and their projected end-of-life dates. As of this week it reads:

After that, the upstream stable maintainers (Greg Kroah-Hartman and Sasha Levin) stop backporting fixes into those two trees. No more 5.10.x or 5.15.x point releases from kernel.org.

Worth knowing: these dates move. It's FOSS reported that the maintainers pushed 6.6 out from December 2026 to December 2027, and 6.12 from December 2026 to December 2028, after what the commit described as "lots of discussions with different companies and groups". kernel.org itself says the projected dates "are not set in stone" and that a new longterm kernel usually starts with a two-year window that can be extended "if there is enough interest from the industry at large". The same article noted 5.10 and 5.15 were still down for December 2026, and that's what the releases page still shows. I wouldn't bank on a last-minute reprieve for kernels that are five and six years old.

Why it probably doesn't matter for your desktop

This is the bit most write-ups skip. Unless you compiled your own kernel from kernel.org, you're not running kernel.org's kernel. You're running your distro's. kernel.org says so directly on that same page: "Unless you downloaded, compiled and installed your own version of kernel from kernel.org, you are running a distribution kernel", and it points you at your distro's support channels rather than its own.

Distros maintain their own kernel packages on their own schedule, and that schedule is the one that matters for you. The clearest example is Ubuntu 22.04 LTS. Its release notes say the server install uses "a non-rolling LTS kernel, v5.15". Canonical's release cycle page lists standard security maintenance for 22.04 running to May 2027, with Expanded Security Maintenance (part of Ubuntu Pro) to May 2032. So a 22.04 server on its stock 5.15 kernel doesn't stop getting kernel fixes in December just because kernel.org does. Canonical carries it.

The same logic applies to Debian, RHEL and friends: read your distro's lifecycle page, not kernel.org's table. If your distro release is still supported and you're installing updates, you're fine.

The flip side: if the distro release itself is out of support, the kernel date is the least of it. Debian 11, for example, reached the end of its LTS on 31 August 2026, and Debian says it will not provide further security updates for it from September (a subset of packages is supported by external parties under Extended LTS). A bullseye box needs a release upgrade, not a kernel swap.

The machines that do care

Where December actually bites is anything running a kernel that someone else built once and stopped updating. Upstream 5.10 and 5.15 were the base for a lot of embedded and vendor kernels, and when upstream stops, the vendor's fixes often stop with it unless they're doing the backporting themselves. It's FOSS makes the same point: on embedded systems, single-board computers and servers, "moving across kernel versions is tricky, and EOL dates are a big deal."

In a typical homelab, that's things like:

None of these is anyone's fault, by the way. Hardware gets certified against a kernel, the kernel ages out, and the vendor has to decide whether a rebase is worth the work. That's just how the economics of long-lived devices go. The industrial world deals with it through the Linux Foundation's Civil Infrastructure Platform, which maintains "Super-Long-Term Stable (SLTS) kernels with a maintenance period of ten years" for things like railways and power grids. Your Pi-hole box is not getting that treatment, and that's fine. It just means you're the maintenance team.

The five-minute check

On each Linux machine you look after, run two commands. Both are read-only.

uname -r
cat /etc/os-release

uname -r is the check kernel.org itself suggests. Its rule of thumb: "If you see anything at all after the dash, you are running a distribution kernel." Something like 5.15.0-xxx-generic is Ubuntu's 5.15, maintained by Canonical. A bare 5.15.180 with nothing after it is more likely something compiled straight from upstream, and that's the one to look at harder.

/etc/os-release is the standard identification file described in the systemd os-release docs (systemd 262 at time of writing). Look at PRETTY_NAME and VERSION_ID, then check that release against your distro's lifecycle page.

That gives you three buckets:

  1. Supported distro, distro kernel. Install updates as normal. Done.
  2. Unsupported distro release. Plan the release upgrade. The kernel comes along for the ride.
  3. Vendor or self-built kernel on 5.10 or 5.15. Check whether the vendor has a newer image or firmware. If not, decide whether the box is exposed to anything that matters.

For bucket three on an Ubuntu 22.04 machine, there's also the option of moving to the Hardware Enablement (HWE) kernel, which Canonical's kernel lifecycle page documents with:

sudo apt-get install --install-recommends linux-generic-hwe-22.04

I haven't run that against every flavour of 22.04, and a stock 22.04 server on the GA kernel is already covered by Canonical, so I'd only bother if you want newer hardware support.

What I'd actually do

My honest take: don't treat December as a deadline for everything. Treat it as the excuse to do the audit you've been meaning to do. Most of your machines will land in bucket one and you'll have spent thirty seconds on each.

For the leftovers, the boring option is usually right. If a box only lives on your LAN, does one job and has no ports open to the internet, an unpatched kernel is a risk you can weigh and accept for a while. If it's exposed, reflash it with a current image or retire it. What I wouldn't do is leave it in an unknown state, because "I'm not sure what kernel that runs" is exactly the box that bites you later.

If you only remember one thing: kernel.org's date is for kernel.org's kernels. Check your distro's date, and save your worry for the devices that don't have one.