{"id":"CVE-2026-94624","published":"2026-09-21T22:17:01.280","lastModified":"2026-09-21T22:17:01.280","description":"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.","cvssScore":7.5,"cvssSeverity":"HIGH","cvssVector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","cwes":["CWE-770"],"vendors":[],"products":[],"references":[{"url":"https://github.com/vllm-project/vllm","tags":[]},{"url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/v1/kv_offload/tiering/p2p/control/zmq.py#L240-L260","tags":[]},{"url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/v1/kv_offload/tiering/p2p/manager.py#L667-L677","tags":[]},{"url":"https://github.com/vllm-project/vllm/pull/51504","tags":[]},{"url":"https://www.vulncheck.com/advisories/vllm-through-0.29.0-denial-of-service-via-unbounded-p2p-kv-offloading-sessions","tags":[]}],"exploitRefs":[{"url":"https://github.com/vllm-project/vllm","tags":[]},{"url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/v1/kv_offload/tiering/p2p/control/zmq.py#L240-L260","tags":[]},{"url":"https://github.com/vllm-project/vllm/blob/v0.29.0/vllm/v1/kv_offload/tiering/p2p/manager.py#L667-L677","tags":[]},{"url":"https://github.com/vllm-project/vllm/pull/51504","tags":[]}],"hasPoc":true,"ai":{"summary":"The vulnerability allows attackers to create unreachable peer sessions that exhaust ZeroMQ socket quotas, leading to EngineCore crashes and service disruptions.","exploitability":"Exploitation requires configuring OffloadingConnector with TieringOffloadingSpec and supplying specific remote host and port values; the attack is moderately complex.","blast_radius":"If exploited, this could result in widespread service outages affecting all inference processes relying on P2P KV offloading.","remediation":"Update to a patched version of vLLM or disable P2P KV offloading until a fix is available.","tags":["dos","p2p","kv-offloading","zeromq"],"model":"qwen2.5:7b-instruct","analyzedAt":"2026-09-22T06:15:07.196Z"}}