Blog

nginx2root: Check Every nginx You Didn't Know You Had

Critical nginx HTTP/2 flaws may allow remote code execution. Update nginx and check every panel, container, appliance, and platform that bundles it.

Dmytro
nginx nginx2root cve-2026-123007 cve-2026-123633

nginx2root: Check Every nginx You Didn’t Know You Had

Apparently, Black October looked at the calendar, shrugged, and continued into Black November.

Two newly disclosed vulnerabilities in nginx can allow an unauthenticated attacker to execute code through HTTP/2. Fixed releases are available, public proof-of-concept code exists, and nginx is installed in far more places than the server where you remember typing apt install nginx.

The immediate task is to find every copy, determine how it is maintained, and update it. That includes nginx installed directly on a server, supplied by a hosting control panel, embedded in an appliance, or hiding inside a Docker image with a reassuringly green status indicator.

What happened

The main vulnerability is CVE-2026-123007, also called nginx2root. It is an integer underflow in nginx HTTP/2 flow control that can lead to an out-of-bounds write in memory. HTTP/2 only needs to be enabled on a reachable listener for the vulnerable code to be exposed.

The bug has been present since nginx 1.9.5, released in 2015. Ten years is an impressively long time for a bug to sit quietly in one of the internet’s busiest waiting rooms.

A second HTTP/2 flaw, CVE-2026-123633, is a use-after-free condition. It can be reached when nginx proxies requests to an upstream server and an active HTTP/2 connection is closed during connection draining under specific resource conditions.

Both can result in code execution with the privileges of the nginx worker. That does not automatically provide root access to the entire host, but a compromised web-server process can expose proxied applications, credentials available to that process, configuration data, and traffic. It can also provide a useful starting point for exploiting another vulnerability or configuration mistake.

Wiz researchers demonstrated remote code execution, including a bypass of address space layout randomization. Other researchers have published technical demonstrations and proof-of-concept code. At the time of publication, the vendor and researchers had not reported exploitation in the wild. We would prefer to keep it that way.

Update to a fixed build

For nginx Open Source, the fixed upstream versions are:

  • nginx 1.30.7 on the stable branch
  • nginx 1.31.9 on the mainline branch

CVE-2026-123007 affects nginx Open Source versions from 1.9.5 through 1.31.8, including stable releases through 1.30.6. The new releases also fix several other security problems affecting features such as HTTP/3, DNS resolution, TLS request processing, mail proxying, try_files, and stream modules. Even systems without HTTP/2 may therefore need the update.

NGINX Plus has its own fixed builds and version-specific guidance. Follow the F5 advisory for the exact release applicable to your installation.

If nginx comes from Debian, Ubuntu, AlmaLinux, another operating system, or a commercial repository, install the package identified by that vendor’s security bulletin. Distribution maintainers often backport fixes without changing the visible upstream version in the way you might expect. nginx -v alone may therefore be insufficient. Check the full package release against the vendor advisory.

After installing the update, perform a full restart or the supported live binary upgrade and verify the running processes. A configuration reload alone can leave the old executable and old workers serving existing connections.

Before restarting, naturally, test the configuration:

nginx -t

This command checks nginx syntax. It does not test your application, upstream services, certificates, firewall, DNS, or whether Monday has developed opinions about your maintenance window.

nginx is often part of something else

The most important part of this incident may be inventory rather than the command used to update one obvious web server.

Vesta and Hestia installations commonly use nginx in front of Apache or PHP-FPM. Plesk can run nginx as a reverse proxy. CloudPanel uses nginx as its web server, while aaPanel can install and manage it as part of its stack. Nginx Proxy Manager puts a friendly interface around nginx and is frequently deployed as a container. Kubernetes ingress products, API gateways, NAS platforms, mail systems, monitoring packages, and vendor appliances may also include nginx internally.

This does not mean every product using nginx is automatically vulnerable. It may contain a patched build, exclude the affected module, disable HTTP/2, or follow a separate release cycle. It does mean that the product must be checked.

For a control panel or appliance, start with the product vendor’s advisory and supported update mechanism. Replacing its nginx binary manually can break generated configurations, dependencies, or future updates. Update the complete product if that is what the vendor requires.

For Docker and other container deployments, identify the image that actually contains nginx. Pull or build a corrected image, redeploy it, and remove old replicas after verification. Updating nginx on the container host does not alter a binary sealed inside an existing image. The same applies to an ingress controller: update the controller image or chart according to its maintainer’s guidance, not merely the host operating system.

Useful places to check include:

  • Public web and reverse-proxy servers
  • Vesta, Hestia, Plesk, CloudPanel, aaPanel, and similar hosting panels
  • Nginx Proxy Manager and other Docker or Compose stacks
  • Kubernetes ingress controllers and gateway products
  • NAS devices, security appliances, monitoring systems, and vendor applications
  • Internal services, staging systems, old VDS instances, and forgotten test environments

Do not rely on the HTTP Server header as inventory. It can be hidden, changed, supplied by a CDN, or describe only the outermost proxy. Search package databases, processes, container images, Compose files, Kubernetes manifests, listening sockets, and product documentation.

If you cannot update immediately

Disabling HTTP/2 on exposed listeners temporarily blocks the paths used by CVE-2026-123007 and CVE-2026-123633. Depending on the nginx version and configuration style, this means removing http2 from listen directives or setting:

http2 off;

Test the configuration, restart nginx, and confirm from outside that the listener no longer negotiates HTTP/2. Remember that this does not mitigate the other vulnerabilities fixed in the same nginx releases, including issues in HTTP/3, resolver, and mail-proxy functionality. It is a short bridge to an update, not a new long-term configuration policy.

If a CDN sits in front of nginx, test and update the origin too. A CDN may terminate HTTP/2 before forwarding HTTP/1.1 to the origin, which can reduce exposure to these two flaws, but it does not prove that the origin is unreachable directly or that other newly fixed nginx issues do not apply.

Prepare the update before pressing Enter

This needs prompt attention, but prompt does not have to mean improvised. Schedule the work in the next available maintenance window, confirm that backups exist, and make sure they can actually be restored. Back up application data, databases, certificates, relevant configuration, and the information required to rebuild the service. Store those copies independently from the server being updated.

Have a rollback and continuity plan. For an important service, that may include an additional VDS in another location, already prepared to receive the workload if the update reveals a compatibility problem. An ITLDC NVMe VDS can be provisioned quickly, but “quickly” is still much nicer when deployment scripts, data replication, DNS changes, certificates, secrets, and health checks have been considered before the primary server stops answering.

A useful plan B is more than an empty virtual machine. Know how current data reaches it, how traffic will be switched, how long DNS or load-balancer changes take, and how you will verify the result. Then test the procedure occasionally. Untested disaster recovery is mostly optimistic documentation.

Find it, update it, verify it

The nginx HTTP/2 flaws deserve attention because they are remotely reachable without authentication, affect versions going back a decade, and sit inside software used at the front door of countless applications. Patches are available, so this is not a situation that benefits from prolonged philosophical discussion.

Inventory every nginx deployment, including the copies supplied by panels, containers, ingress controllers, appliances, and internal products. Install the vendor-approved update, restart the actual running service, and verify the version and application behavior afterward.

Then keep the inventory. At the rate October and November are going, it may save time before the calendar discovers December.

Need Help?

Our support team is available 24/7 to assist you with any questions or issues.

Contact Support