---
id: "GHSA-28qf-cpwq-vmc2"
page: "https://infrastack.ai/live/GHSA-28qf-cpwq-vmc2"
kind: "vulnerability"
aliases: ["GHSA-28qf-cpwq-vmc2"]
title: "Pure-Python parser request replay and buffer growth behind a declined upgrade"
severity: "medium"
cvss: null
epss: null
exploited: false
status: "open"
category: "package"
products: [{"name": "aiohttp", "fixed": ["3.14.4"], "affected": "<= 3.14.3", "ecosystem": "pypi"}]
ecosystems: ["pypi"]
tags: ["aiohttp", "cwe-770", "cwe-835", "package", "python", "web-servers"]
not_affected_if: ["c-parser", "standard-wheel"]
published_at: "2026-11-08T18:17:48+00:00"
first_seen_at: "2026-11-09T22:53:22+00:00"
updated_at: "2026-11-09T22:58:37+00:00"
sources: [{"source": "ghsa-repo", "url": "https://github.com/aio-libs/aiohttp/security/advisories/GHSA-28qf-cpwq-vmc2", "first_seen_at": "2026-11-09T22:53:22+00:00"}]
exploit_sources: []
---

# Pure-Python parser request replay and buffer growth behind a declined upgrade

Link for people: https://infrastack.ai/live/GHSA-28qf-cpwq-vmc2 (this .md is for agents)

When the pipelined request queue is full, the pure-Python HTTP parser keeps the unparsed remainder and also returns it to the caller, so after a declined upgrade the data is re-fed. An attacker can make the buffer grow with each request and get already-handled requests dispatched to the application again.

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

## Affected

- pypi:aiohttp affected <= 3.14.3 fixed 3.14.4

## How to check

- Run `python -c "import aiohttp; print(aiohttp.__version__)"` and confirm it is `3.14.3` or lower.
- Check whether the C extensions are present: `python -c "from aiohttp import http_parser; print(http_parser.HttpRequestParser)"` and look for a pure-Python class instead of the C one.
- Check whether `AIOHTTP_NO_EXTENSIONS` is set in the environment or deployment config.

## Fix

- `aiohttp` `3.14.4`

## Timeline

- 2026-11-09T22:53:22+00:00 ghsa-repo https://github.com/aio-libs/aiohttp/security/advisories/GHSA-28qf-cpwq-vmc2

## Documents

- [advisory] Maintainer's advisory: github.com https://github.com/aio-libs/aiohttp/security/advisories/GHSA-28qf-cpwq-vmc2

## From the sources

### Summary

When the pipelined request queue was full, the pure-Python parser stored the unparsed remainder internally and also returned it to the caller. After a handler declined an upgrade request, the buffered data was re-fed, so the buffer grew with each handled request and already-handled requests were dispatched to the application again.

### Impact

If users are using the Python parser (`AIOHTTP_NO_EXTENSIONS`, or not a standard wheel installed), an attacker could cause excessive memory use and cause pipelined requests to be processed more than once.

### Workaround

Switch to the default C parser.

---

Patch: https://github.com/aio-libs/aiohttp/commit/47babd8c23ca79e2bae6cc8dcb5752b61e369fc7

---

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