Skip to content
CVE-2026-123007: Update NGINX to a fixed release

CVE-2026-123007: Update NGINX to a fixed release

November 19, 2026

NGINX identifies CVE-2026-123007 as memory corruption in HTTP/2 flow control and rates it major. Upgrade affected deployments to a fixed release through their supported package or image channel. The NGINX security advisory list identifies the issue; the November 18, 2026 release announcement confirms the stable and mainline fixes.

Choose a fixed build for your branch

Upstream branchFixed release for this CVE
Stable 1.301.30.7 or later in that branch
Mainline 1.311.31.9 or later in that branch

The advisory lists an affected upstream span of 1.9.5 through 1.31.8, alongside explicit fixed-release exceptions. Read those together: stable 1.30.7 is fixed even though its version sorts below mainline 1.31.8. A rule that flags every version below 1.31.9 would incorrectly reject the stable fix.

For distribution packages, record the complete package revision and consult the distributor’s security status for this CVE. Vendors can backport fixes without changing the upstream-looking version. For NGINX Plus or a bundled appliance, use the product vendor’s applicable fixed build; the Open Source table is not a product-release mapping.

Identify the deployments that need attention

Inventory every NGINX instance, including replicas, container images and standby nodes. Check which binary or image the service actually starts, its package origin, and whether HTTP/2 is enabled in the effective configuration. The HTTP/2 module documentation explains build support and protocol configuration. Build support alone does not establish that a listener uses HTTP/2.

Inspect the installed binary’s version and build information using the documented version options. Review configuration through your normal administrative access, including included files and server-level settings. Configuration dumps can contain credentials or internal addresses; keep them in an approved private location. The command-line reference distinguishes version reporting, configuration testing and configuration dumping.

Reduce exposure while arranging the update

If a fixed build cannot be deployed immediately, consider temporarily disabling HTTP/2 on the affected NGINX listeners, provided the service can operate without it. This reduces use of the affected protocol; it does not install the fix. Clients or upstream services that require HTTP/2 need a tested alternative before this change.

Use the configuration syntax supported by the installed version. The standalone HTTP/2 directive was introduced in 1.25.1, so it is not a universal instruction for every affected release. Check all relevant server blocks, validate the proposed configuration, and apply it through the service’s normal change process. Confirm ordinary clients negotiate the intended protocol afterward. Do not treat a change on one virtual host as evidence that every listener is covered.

Verify that the running service was replaced

Deploy a fixed package or image using the vendor’s upgrade procedure. Validate configuration before the rollout, update a small portion of the fleet first where possible, and check normal health, routing and TLS behavior before proceeding. Preserve a recovery plan that does not silently restore an exposed build.

A successful package installation or an updated image tag does not prove that running processes changed. Check the deployed image digest or package revision and the service’s running executable on each instance. Complete the restart or supported binary-upgrade procedure, and confirm that old processes have drained and exited. NGINX documents configuration reloads and executable upgrades separately; a configuration reload alone is not proof of a new executable.

Finish with ordinary application requests and operational checks: healthy replicas, expected protocol negotiation, normal error rates and no unexpected worker restarts. These checks assess rollout health, not exploitability. Keep the fixed build identifier and rollout result in the change record, and revisit any temporary protocol restriction once the update is confirmed.

Acknowledgements

This post was created with assistance from AI.

Last updated on