Skip to main content
Apache 2.4.71: the HTTP/2 patch that shouldn't wait until Monday

This post is a translation of the Spanish original.

3 min read

Apache 2.4.71: the HTTP/2 patch that shouldn't wait until Monday

The first thing I did when I read the advisory was ask myself how many Apache servers I know with HTTP/2 enabled without anyone consciously deciding it. Probably more than I'd like.

On 11 November Apache HTTP Server 2.4.71 came out, fixing CVE-2026-119618: a race condition in mod_http2's assign_headers() function that can corrupt heap memory and end in remote code execution. Without authentication. Apache rates it critical and Red Hat gives it a CVSS of 9.8.

What is affected and what isn't

It affects versions 2.4.55 to 2.4.70, and also the standalone mod_h2 module from 2.0.8 to 2.0.42. The key condition is that HTTP/2 is actually in use: mod_http2 loaded and h2 or h2c in the Protocols directive. If your server only speaks HTTP/1.1, this flaw doesn't touch you.

Checking takes a minute:

bash
# Is the module loaded? (on Debian/Ubuntu, apache2ctl)
apachectl -M 2>/dev/null | grep -i http2

# Is HTTP/2 advertised in the configuration?
grep -Ri "^\s*Protocols" /etc/apache2/ /etc/httpd/ 2>/dev/null

If you can't update yet, Red Hat suggests removing h2 and h2c from Protocols to serve only HTTP/1.1 in the meantime. Watch one detail: if Apache sits behind another Apache acting as a reverse proxy, the one facing the world needs the patch too.

On whether there's an exploit, sources disagree. Red Hat mentions public research that demonstrated reliable exploitation with the default configuration and mod_http2 enabled; other sources say there's no linked public code or known exploitation, and as of 12 November it wasn't in CISA's KEV catalogue. I don't take it as confirmed, but I wouldn't relax because of that either.

Who found it, and why it matters to me

The oss-security announcement credits the discovery to Anthropic's open-source scanner and to Anthony Weems and Simon Scannell of Google Cloud Vulnerability Research. An AI-powered tool appearing in the credits of a critical RCE in the internet's most veteran web server is no longer as surprising as it was a year ago. And that's what worries me.

If finding race conditions in mature C code becomes cheaper, it becomes cheaper for everyone, not just for those who report responsibly. The time between a patch being published and someone diffing it against the previous version to build the exploit gets shorter.

My take: this gets patched today, not in the next window

I've been in systems for years and I understand the opposite argument perfectly. Production changes go through change control, some applications break with an update and nobody wants to be the one who took the website down on a Friday afternoon. It's a serious argument and it's almost always right.

But with a pre-authentication flaw, in an internet-facing service and with a trivial mitigation, I think the monthly window is a bad habit. If you can't deploy 2.4.71 or your distribution's backport today, at least disable HTTP/2 today. Losing multiplexing for a few days is far less serious than losing the server.

What unsettles me most isn't this particular bug, it's the pace. If AI analysis tools keep finding flaws of this calibre in projects we considered stable, are we in systems ready to patch in hours what we used to patch in weeks?

Sources

David González Lombraña

Apuntes, laboratorio y proyectos de IA, Linux y redes.