Skip to main content
PathMon
Back to blog
SecurityPatch Management

NGINX2Root (CVE-2026-123007): Patch Your nginx Fleet

NGINX2Root is fixed in nginx 1.31.9 and 1.30.7. What Ubuntu, Debian, RHEL, Alpine and Amazon Linux shipped, and how to prove every server is patched.

By PatchMon Team

Last updated: 19 November 2026. Distro status in this post was checked against each vendor's own tracker at 07:20 UTC on 19 November 2026. Debian, Red Hat and SUSE have not shipped fixes yet, so we will keep the tables below current as they do.

If you run a Linux server that talks to the internet, there is a good chance nginx is on it. It might be the web server, the reverse proxy in front of an app, the ingress controller in a Kubernetes cluster, or the thing a vendor appliance quietly bundled.

On 18 November 2026 the nginx project released nginx 1.31.9 and 1.30.7, fixing ten vulnerabilities. Five of them, led by CVE-2026-123007, are what researchers at Xint have named NGINX2Root: memory corruption bugs that an unauthenticated attacker can reach over the network, each of which Xint says leads to remote code execution. Xint then chained them with a separate, still unpatched Linux kernel bug to get root on the host, including from inside a Docker container.

The fix is out. The question for most teams is how many of their nginx processes are still running the old code.

The Short Version

If you only have two minutes:

  • What it is: Five memory corruption bugs in nginx: HTTP/2 flow control (CVE-2026-123007), HTTP/2 connection handling (CVE-2026-123633), the resolver (CVE-2026-123634), HTTP/3 (CVE-2026-121810) and the mail proxy (CVE-2026-113425).
  • Severity: F5, the CVE numbering authority for nginx, scores all five CVSS v4.0 9.2 (CVSS v3.1 8.1). nginx's own advisory rates CVE-2026-123007 and CVE-2026-113425 "major" and the other three "medium".
  • Who is affected: Any nginx from 1.9.5 to 1.31.8 with HTTP/2 enabled is exposed to the lead bug. The others need specific features: HTTP/2 proxying, request-time DNS with subrequests, HTTP/3, or the mail proxy.
  • Fixed in: nginx 1.31.9 (mainline) and 1.30.7 (stable). NGINX Plus 37.0.8.1 and 37.3.0.1.
  • Exploitation so far: No reported exploitation in the wild, and none of the five is in CISA's KEV catalogue as of 19 November. A public proof of concept exists but only works in lab conditions.
  • What you do about it: Install your distro's or nginx.org's fixed package, then restart nginx. A reload does not replace the running binary.

What NGINX2Root Actually Is

Xint's technical write-up describes five independent attack paths, not a chain: "Each one of the five nginx vulnerabilities are independent attack paths and do not need to be chained together. They differ in what configuration is affected."

CVE-2026-123007, HTTP/2 flow control. This is the one that matters to almost everyone. In the words of the CVE record, "the flow control window can exceed the allowed threshold, potentially resulting in a heap buffer overflow typically followed by an NGINX worker segmentation fault." The record adds that attackers "can execute code on systems with Address Space Layout Randomization (ASLR) disabled or when the attacker can bypass ASLR." The only precondition is HTTP/2 being turned on, and the bug has been there since nginx first shipped HTTP/2 in 1.9.5.

The other four each need a particular configuration:

CVE Where Reachable when nginx severity Vulnerable versions
CVE-2026-123007 HTTP/2 flow control HTTP/2 enabled Major 1.9.5 to 1.31.8
CVE-2026-123633 HTTP/2 HTTP/2 clients proxied to a backend, worker low on worker_connections Medium 1.19.7 to 1.31.8
CVE-2026-123634 Resolver proxy_pass with variables, plus SSI or mirror subrequests Medium 0.6.18 to 1.31.8
CVE-2026-121810 HTTP/3 HTTP/3 (QUIC) enabled Medium 1.25.0 to 1.31.8
CVE-2026-113425 Mail proxy nginx configured as a mail proxy Major 0.5.33 to 1.31.8

Sources: nginx security advisories for severity and versions, the F5-published CVE records for the conditions.

Where "2Root" comes from

On its own, each nginx bug gives an attacker code execution as the nginx worker user, usually www-data or nginx. Xint found a second bug in the Linux kernel's SCTP code, a use-after-free in the send path that it says was "introduced in Linux v4.15 in 2018". With it, "an attacker already running code as an nginx worker could reliably gain root." Xint also "obtained a root shell on the host in a setup where nginx was running in a Docker container."

As of 18 November, the kernel bug has no CVE and no fix, and Xint is holding back its details until there is one. What it has said is the part that matters for defenders: "On many Linux distributions, the SCTP module is installed by default and the kernel loads it automatically the first time a program opens an SCTP socket." A fully patched kernel does not close this. See the mitigations below.

Why This One Hurts More Than the Usual nginx CVE

nginx has had an unusually busy year for advisories, and most of them were fixed on a normal patch cycle without drama. Xint counted 26 by 17 November, against two in all of 2025. Three things make this release different.

It is unauthenticated and needs nothing but HTTP/2. Many of this year's nginx advisories needed a specific directive combination, such as alias with try_files, or map with a regex. CVE-2026-123007 needs only HTTP/2 to be enabled. Xint calls an unauthenticated nginx RCE "a holy grail for security researchers" and points to CVE-2013-2028 as one of the very few previous examples.

The exposed population is enormous. From a Shodan snapshot on 10 November, Xint found 56.7 million hosts identifying as nginx. Of the 11.94 million that reported a version, 98.9% fell inside the vulnerable range. Xint is careful to say that version strings do not prove exposure, since HTTP/2 must be enabled and many distro builds carry backported fixes, but the scale is clear.

Several teams found it independently, with AI help. The CVE record for CVE-2026-123007 credits four separate reporters: Rapid7, Wiz Research, Xint, and Ikotas Labs "using GPT-6 Astra". Xint says a "fully automated Xint scan" found the other four bugs. When four groups find the same bug at once, it is reasonable to assume others have too.

Exploitation So Far

Here is what is solid as of 19 November:

  • No reported exploitation. None of the five CVEs is in CISA's KEV catalogue (catalogue version 2026.11.18). Tenable reports "no reports of threat actors exploiting these CVEs as of November 18."
  • A public exploit exists, and it is fragile. Rapid7's Stephen Fewer published a proof of concept for CVE-2026-123007. Its README says it "is only suitable to demonstrate the given RCE exploitation strategy against a local freshly started Docker container" and "will likely fail against a vulnerable NGINX server which has multiple concurrent clients". Rapid7's analysis explains why: the attack depends on a specific x86-64 glibc memcpy path and a large volume of HTTP/2 traffic.
  • Full exploits are being withheld. Xint says it "intentionally omit[s] the final steps needed to reproduce the full exploits."
  • Be careful with GitHub. Around twenty repositories each for CVE-2026-123007 and nginx2root appeared within a day. Fake proof-of-concept repositories carrying malware are a familiar pattern after high-profile CVEs. Do not run any of them on a machine you care about.

Fragile exploits improve. Patch as if exploitation were confirmed.

Which nginx Versions Are Fixed

Product Fixed release
nginx mainline 1.31.9
nginx stable 1.30.7
NGINX Plus 37.0.8.1 and 37.3.0.1. R33 and later R releases get no fix: move to 37.x
nginx 1.29.x and older No upstream release. Get the fix from your distro, or upgrade

Both open source releases came out on 18 November (nginx.org). The same releases also fix five further bugs that are not part of NGINX2Root, including CVE-2026-118899, an infinite loop when reading a request body over TLS 1.3 that nginx rates "major". Distro advisories mostly cover all of them together.

What Your Distro Is Shipping

Most Linux servers do not run nginx from nginx.org. They run their distribution's package, and distributions backport fixes without changing the upstream version number. As Xint puts it, "A backport can fix a vulnerability without changing the version reported by nginx -v." A Ubuntu 24.04 host reporting nginx/1.24.0 can be fully patched. What matters is the full package version.

Distribution Package Fixed version Advisory
Ubuntu 26.10 nginx 1.30.4-7ubuntu1.2 USN-9159-1, 19 Nov
Ubuntu 26.04 LTS nginx 1.28.3-2ubuntu1.13 USN-9159-1
Ubuntu 24.04 LTS nginx 1.24.0-2ubuntu7.20 USN-9159-1
Ubuntu 22.04 LTS nginx 1.18.0-6ubuntu14.23 USN-9159-1
Ubuntu 20.04 and older nginx ESM builds, Ubuntu Pro only USN-9159-1
Debian 13 "trixie" nginx Not fixed yet Debian tracker
Debian 12 "bookworm" nginx Not fixed yet Debian tracker
RHEL 8, 9, 10 nginx, nginx:1.24, nginx:1.26 Not fixed yet (Affected) Red Hat CVE page
AlmaLinux 10 nginx 1.26.3-16.el10_3.1.alma.2 ALSA-2026:A024
AlmaLinux 9 nginx 1.20.1-32.el9_9.1.alma.3 ALSA-2026:A025
Amazon Linux 2023 nginx 1.30.7-1.amzn2023.0.1 ALAS2023-2026-3494
Amazon Linux 2 (nginx1 extra) nginx 1.30.7-1.amzn2.0.1 ALAS2NGINX1-2026-018
Alpine edge, 3.24 nginx 1.30.7-r0 Alpine security tracker
Alpine 3.23, 3.22 nginx 1.28.3-r8 Alpine secdb
SUSE, openSUSE nginx Not fixed yet (under analysis) SUSE CVE page
FreeBSD nginx, nginx-devel 1.30.7,3 and 1.31.9 VuXML

Ubuntu

Canonical shipped USN-9159-1 on 19 November, covering every supported release and eight of the ten CVEs in this nginx release, including all five NGINX2Root bugs. The upstream version strings stay at 1.18.0, 1.24.0 and so on; only the Ubuntu revision changes. Releases from 20.04 back are covered through Ubuntu Pro (ESM).

Debian

As of 19 November, Debian's tracker lists bookworm and trixie as vulnerable, including the current security archive builds (1.22.1-9+deb12u10 and 1.26.3-3+deb13u10). Only unstable has a fix so far, in 1.30.6-4. Watch the tracker page for a DSA, and apply the mitigations below until it lands.

RHEL, Rocky Linux and AlmaLinux

Red Hat rates CVE-2026-123007 Critical and lists RHEL 8, 9 and 10 as affected across the nginx package and the nginx:1.24 and nginx:1.26 module streams, with no advisory yet. Check which stream each host has enabled with dnf module list --enabled nginx. AlmaLinux did not wait and published its own fixes for AlmaLinux 9 and 10 (ALSA-2026:A025 and ALSA-2026:A024) on 18 November. Rocky Linux rebuilds Red Hat's packages, so expect its fix to follow Red Hat's.

Amazon Linux

Amazon Linux 2023 moved straight to upstream 1.30.7 in ALAS2023-2026-3494. On Amazon Linux 2, nginx comes from the nginx1 extras topic, fixed in ALAS2NGINX1-2026-018.

Alpine and containers

Alpine edge and 3.24 carry upstream 1.30.7-r0, and 3.22 and 3.23 have the fixes backported into 1.28.3-r8 (per Alpine's security database). The official nginx Docker image was rebuilt on release day: the mainline and 1.31.9 tags on 18 November, the stable and 1.30.7 tags early on 19 November. Pull the new image, rebuild anything that uses it as a base, and redeploy. A running container does not pick up a new image on its own.

Kubernetes ingress-nginx and OpenResty

The community ingress-nginx controller was archived in March 2026 and will not get a fix. If you still run it, disable HTTP/2 with use-http2: "false" in its ConfigMap and plan the move off it. F5's own NGINX Ingress Controller is a separate product with its own releases. OpenResty's latest tag as of 19 November is v1.31.8.1, built on the vulnerable nginx 1.31.8.

If You Cannot Patch Right Now

The CVE record for CVE-2026-123007 lists the workaround as "None". These are holding measures from Xint and Red Hat for the hours or days before a fixed package is installed. Each one covers only some of the bugs.

Check what you are exposed to:

# Installed nginx packages and versions
dpkg -l 'nginx*' | awk '/^ii/ {print $2, $3}'        # Debian / Ubuntu
rpm -qa 'nginx*' --qf '%{NAME} %{VERSION}-%{RELEASE}\n'  # RHEL family / Amazon / SUSE

# Is HTTP/2 enabled? (reads the config on disk)
sudo nginx -T 2>&1 | sed 's/#.*//' \
  | grep -nE 'listen[[:space:]].*http2|^[[:space:]]*http2[[:space:]]+on[[:space:]]*;'

# HTTP/3, resolver, mirror and mail usage
sudo nginx -T 2>&1 | sed 's/#.*//' | grep -nE 'quic|http3|resolver|mirror|^[[:space:]]*mail'

Disable HTTP/2 (blocks CVE-2026-123007 and CVE-2026-123633). On nginx 1.25.1 and later, set http2 off; in every affected server block and remove http2 from the listen lines. On older versions, just remove http2 from listen:

listen 443 ssl;
http2 off;

Clients fall back to HTTP/1.1, but anything that needs HTTP/2, such as native gRPC, will break.

Disable HTTP/3 (blocks CVE-2026-121810). Remove quic listeners, set http3 off; and add add_header Alt-Svc 'clear' always;.

Remove request-time DNS with subrequests (CVE-2026-123634). Point proxy_pass at a fixed upstream group rather than a variable, and turn mirror off; where you do not need it.

Lock down the mail proxy (CVE-2026-113425). Bind mail listeners to loopback or allow only trusted networks.

Restart, do not just reload. Xint notes that a reload "applies the new configuration only to new connections", and existing HTTP/2 sessions stay on the old workers. After a config change for mitigation, check for nginx: worker process is shutting down in ps and restart if they linger.

Block the SCTP module on hosts that do not use SCTP. This is the only defence against the root escalation until a kernel fix exists:

echo "install sctp /bin/false" | sudo tee /etc/modprobe.d/nginx2root-sctp.conf
lsmod | grep sctp   # if this prints anything, SCTP is already loaded

Xint is explicit about the limits: this "does not unload a module that is already loaded, and it does not disable SCTP compiled into the kernel." If lsmod shows sctp, plan a reboot after adding the file.

Watch what nginx spawns. Xint observes that "a normal nginx worker has no reason to run external programs." A shell or download tool started by the nginx worker user is the clearest sign of a successful attack. Repeated worker process ... exited on signal 11 lines in error.log can indicate failed attempts.

How to Tell Your Fleet Is Actually Patched

nginx has the same trap as PHP did with PHP2Shell: a fixed package on disk is not a fixed process in memory. Four things tend to go wrong:

  1. The package updated, the service did not restart. Xint is clear: "A configuration reload alone does not replace the running executable." On Linux you can see it directly. If the binary was replaced underneath a running master process, its executable link shows as deleted:
sudo ls -l /proc/$(cat /run/nginx.pid)/exe
# ... -> /usr/sbin/nginx (deleted)   means the old binary is still running
  1. nginx is somewhere you did not look. The reverse proxy in front of an internal tool, the ingress controller, a container image built from nginx:1.27 six months ago, the appliance with a web UI.
  2. The version string tells you nothing on its own. nginx/1.24.0 on Ubuntu 24.04 says nothing; 1.24.0-2ubuntu7.20 does. The changelog is the definitive answer on Debian-family systems:
apt changelog nginx 2>/dev/null | grep -E 'CVE-2026-123007'
  1. Your distro has not shipped yet. Debian, RHEL and SUSE are the obvious examples this week. "No update available" is not the same as "fixed".

Running those checks on one server takes a minute. Running them across two hundred, and keeping the answer current as Debian and Red Hat ship over the coming days, is the actual job.

This is the problem PatchMon is built for. Each PatchMon agent reports every installed package, its version, and the updates available for it, including whether an available update is a security update. Search the Packages view for nginx and you get every host carrying an nginx package, the version each one has installed, and which still have an update waiting. As Debian and Red Hat release their fixes, they appear there as pending security updates; as each host updates and reports back, it drops off the list. For containers, the Docker integration flags containers whose image has an update available, which is how you catch the stale nginx:* images. Every update is timestamped, which is the evidence your auditor or customer will ask for afterwards.

If you would like to see that against your own servers, have a look at the PatchMon pricing page and start a PatchMon Cloud trial from there. If you would rather self-host, the Community Edition is free and open source.

Frequently Asked Questions

Is Ubuntu 24.04 patched for CVE-2026-123007? Yes, since 19 November. The fixed package is nginx 1.24.0-2ubuntu7.20 from USN-9159-1. nginx -v will still say 1.24.0.

Is Ubuntu 22.04 patched? Yes. The fixed package is 1.18.0-6ubuntu14.23.

Is Debian patched? Not as of 19 November. Bookworm and trixie are listed as vulnerable on Debian's tracker. Use the mitigations above until a DSA is released.

Does RHEL, Rocky Linux or AlmaLinux have a fix? RHEL does not yet; Red Hat lists RHEL 8, 9 and 10 as affected. AlmaLinux 9 and 10 have their own fixes in ALSA-2026:A025 and ALSA-2026:A024.

Am I affected if HTTP/2 is off? Not by CVE-2026-123007 or CVE-2026-123633. You may still be affected by the HTTP/3, resolver or mail proxy bugs if you use those features, so patch anyway.

Is a reload enough after upgrading? No. Restart nginx, or do a supported live binary upgrade. A reload keeps the old executable running.

Does a kernel update fix the root escalation? Not yet. The SCTP kernel bug Xint used has no CVE and no fix as of 18 November. Block the sctp module on hosts that do not need it.

Is it being exploited in the wild? No exploitation has been reported, and none of the five CVEs is in CISA KEV as of 19 November. A lab-only public proof of concept exists.

Bottom Line

NGINX2Root is five remotely reachable memory corruption bugs in the most widely deployed web server on the internet, one of which needs nothing more than HTTP/2. The fix is in nginx 1.31.9 and 1.30.7, Ubuntu, Amazon Linux, Alpine and AlmaLinux have packages out, and Debian, RHEL and SUSE users are still waiting.

What to do today:

  1. Inventory every nginx: hosts, containers, ingress controllers and appliances, with full package versions.
  2. Update wherever a fix exists, then restart nginx. Verify by package version or changelog, not nginx -v.
  3. Where no fix exists yet, disable HTTP/2 (and HTTP/3, mirror or mail proxy where used) until it does.
  4. Block the sctp kernel module on hosts that do not need it, since the root escalation has no patch.
  5. Rebuild and redeploy container images, and move off the retired community ingress-nginx.

The teams that get through events like this calmly are the ones who can answer "which servers still run a vulnerable nginx?" in minutes. If you want to see how PatchMon does that, the pricing page is the place to start a trial. Either way, get nginx patched and restarted first.