---
id: "CVE-2026-116304"
page: "https://infrastack.ai/live/CVE-2026-116304"
kind: "vulnerability"
aliases: ["CVE-2026-116304"]
title: "PostgreSQL: parallel workers' record type handling lets a database user run code"
severity: "high"
cvss: 8.8
epss: null
exploited: false
status: "open"
category: "runtime"
products: []
ecosystems: []
tags: ["database", "postgresql", "runtime", "self-hosted"]
not_affected_if: ["managed-hosting", "no-untrusted-db-users"]
published_at: "2026-11-12T15:17:38+00:00"
first_seen_at: "2026-11-12T15:41:53+00:00"
updated_at: "2026-11-12T15:43:41+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-116304", "first_seen_at": "2026-11-12T15:41:53+00:00"}]
exploit_sources: []
---

# PostgreSQL: parallel workers' record type handling lets a database user run code

Link for people: https://infrastack.ai/live/CVE-2026-116304 (this .md is for agents)

A use-after-free flaw in how PostgreSQL parallel workers share anonymous RECORD type descriptors lets one process free memory another still uses. A database user with login rights can run arbitrary code as the operating system user running the database.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- Run `SELECT version();` in psql and compare against the fixed versions below.
- Check the installed server binary with `postgres --version`.

## Fix

- PostgreSQL `18.7`
- PostgreSQL `17.12`
- PostgreSQL `16.16`
- PostgreSQL `15.20`
- PostgreSQL `14.25`

## Timeline

- 2026-11-12T15:41:53+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-116304

## From the sources

Use after free in PostgreSQL coordination of anonymous RECORD type identifiers among parallel workers allows a database user to execute arbitrary code as the operating system user running the database, via causing one process to free a shared tuple descriptor while another process retains and later uses its address.  Versions before PostgreSQL 18.7, 17.12, 16.16, 15.20, and 14.25 are affected.

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_