Live › vulnerability › GHSA-8qpc-6rxx-4pxf
Go to dashboardA crafted kv_transfer_params request with an arbitrary remote host and port makes vLLM keep ZeroMQ sockets open for unreachable peers until the context quota runs out. The resulting uncaught ZMQError crashes EngineCore and stops all inference.
Summary written from the sources below. Check the originals before acting.
First seen by InfraStack 2026-10-27 14:47 UTC. All times on this page are UTC.
Deployments of vLLM through 0.29.0 that use the OffloadingConnector with a TieringOffloadingSpec and a peer-to-peer secondary tier are exposed, because that path accepts remote host and port values from request parameters. Servers without that offloading configuration are not described as affected. The advisory itself is withdrawn as a duplicate, so the live details sit under GHSA-frvg-qw88-hww9.
| Product | Ecosystem | Affected | Fixed in |
|---|---|---|---|
| vllm | pypi | >= 0, < 0.30.0 | 0.30.0 |
pip show vllm and confirm the installed version is below 0.30.0.0.30.0 or later.Each row is the first time that source carried this signal. The analyst’s rows are its own findings.
This advisory has been withdrawn because it is a duplicate of GHSA-frvg-qw88-hww9. This link is maintained to preserve external references.
vLLM through 0.29.0 contains a denial of service vulnerability in P2P KV offloading when OffloadingConnector is configured with TieringOffloadingSpec and a peer-to-peer secondary tier. Attackers can supply arbitrary remote host and port values in kv_transfer_params to create unreachable peer sessions that retain ZeroMQ sockets until the context quota is exhausted, causing an uncaught ZMQError that crashes EngineCore and stops all inference.
This page as .md · .json · in your agent: get_signal("GHSA-8qpc-6rxx-4pxf")
GitHub Advisory Database, CC-BY 4.0. Posts quoted belong to their authors.