Security backports for 2.541.x, the last LTS on Java 17: looking for feedback

Hi all,

Before I do anything public, I’d like to hear what the community and the security team think.

Some background on me: I work in DevOps/infrastructure, and builds and releases are part of my everyday work. I maintain the Explain Error plugin and the cpp-linter project, which ships static clang-format/clang-tidy binaries.

2.541.3 is the last LTS that runs on Java 17, and about 82,000 reporting controllers (33%) were on Java 17 in February (stats.jenkins.io). The core fixes from the 2026-06-10, 2026-08-05 and 2026-09-02 advisories went into the current LTS only, as expected. Most people should just move to Java 21, as Mark suggested when this came up in April. Some can’t do that quickly, for example, when a new JDK takes months to get approved internally.

So I tried backporting those core fixes to 2.541.3 on my machine. Remoting and Stapler stay on the versions 2.541.3 ships, with their fixes applied on top. The affected tests pass, most of the new security tests fail on stock 2.541.3 and pass with the patches, and it starts with the top 300 plugins the same way the official build does. Nothing is published.

What I have in mind:

  • Builds of 2.541.3 plus core security backports, nothing else.
  • Each release says which SECURITY ids it covers and which it can’t.
  • Free, with checksums, SBOM, and provenance.
  • A fixed end date (around 12 months), so it’s a bridge to the current LTS, not a fork.
  • 2.541.x only. Older lines (Java 11, Java 8) would need plugin backports too, which I can’t do properly.

Two questions:

  1. Security team and board: any concerns, for example, how to name it so nobody mistakes it for an official LTS, or timing around advisories? Happy to share the patches.
  2. If you’re on 2.541.x: would you use this, and what’s keeping you on Java 17?

Thanks!

You also want to provide downports of all the plugins that already require 2.555.x or higher?

Then that company should revise their strategy to approve new software. Java 21 was released 3 years ago.

I don’t see any good reason why you can’t go to Java 21. What was the bigger problem in the past was the switch from Java8 to Java 11, as there are some older platforms where no newer Java than 8 is available.

One practical detail worth settling early: since Remoting and Stapler are patched as well, a reproducible build needs those artifacts available under their own coordinates. Already released versions cannot be overwritten in the repositories, so anyone rebuilding from source would otherwise pull the unpatched libraries back in.

Identification is the other one. Whatever the version string ends up being, version-keyed vulnerability scanners and bug reports need to be able to tell a patched WAR apart from stock 2.541.3, and the release notes should say plainly that Jenkins security advisories and CVE assignments do not cover it.

The recurring cost is what I would weigh most: every future core advisory means re-applying fixes to a line that gets no upstream work, for as long as people run it. Publishing the patch series as a branch with build instructions may capture most of the value for the 33 percent still on Java 17 without taking on distribution and support of binaries.

1 Like

Good point. No, I don’t plan to backport plugins. I checked the plugin security fixes since 2.541.3: 52 of 56 can still be installed on 2.541.3, and 4 need 2.555.3. But yes, this gap will get bigger over time, so I’d only keep it for about one year.

About the JDK approval, that was just my guess, not a real case I had. I’m still on Java 17 myself. As I understand it, upgrading means I also need to upgrade all my agents. So if a controller on the Java 17 LTS could still get security fixes, that would be nice.

Thanks, these are good points, especially the reproducible build part. I’ll keep them in mind. The branch idea makes a lot of sense as a first step.

One thing to consider, the usage stats are from February, so 8 months old. And they don’t cover 2.555.x the first LTS with Java 21 being mandatory.
I assume that today the number of users on Java 17 is much lower

Ah right. The stats only go to February, before 2.555.1. It was already going down (from 39% to 33% from Sep 2025 to Feb 2026), so it’s probably a lot lower.

Quick update: I put the patches up as a branch with build instructions, as suggested: GitHub - keelinfra/jenkins-extended-security at stable-2.541 · GitHub

No binaries for now. It’s only the core fixes from the June to September advisories, plugins aren’t touched. If a ready-made WAR or Docker image would help you, or anything looks wrong, please tell me.

There’s quite a bunch of fixes from todays advisory: Jenkins Security Advisory 2026-10-28

Thanks for the heads-up. All 15 core fixes from it are on the branch now, details in jenkins-extended-security/COVERAGE.md at 03066d770b16414474f4827b88a4b61585f4a849 · keelinfra/jenkins-extended-security · GitHub