Database/AI/ML frameworks & serving
vLLM: off-by-one prompt-token-ID bound lets a request crash the inference engine
Impact
An authenticated caller that can submit pre-tokenized prompt IDs can pass a token ID equal to the vocabulary size, which escapes validation. With tensor-parallel size 1 and a vocabulary that is an exact multiple of 64, VocabParallelEmbedding.forward does not mask the value before F.embedding, so the GPU raises a device-side assertion and the engine process dies. On a shared serving node that takes down the model for every tenant on that replica and leaves the GPU memory to be reclaimed by a restart, not just the offending request. Single malformed request, repeatable at will.
Who can reach it
Any client authenticated to the vLLM OpenAI-compatible API that is allowed to send prompt token IDs. No special privileges beyond normal inference access.
What to do
Upgrade vLLM to 0.32.0 and restart the serving processes; a rolling restart across replicas avoids a full outage. No configuration mitigation is given in the advisory other than blocking pre-tokenized prompt IDs at a gateway.
References
Related entries
- vLLM: request-supplied chat template bypasses trust_request_chat_template and exhausts the serverCVE-2026-115918 · vLLM /v1/responses endpoint (chat_template_kwargs trust policy)Medium
- MLflow: CrewAI autologging writes provider API keys in plaintext into tracesCVE-2026-116691 · MLflow CrewAI autologging (mlflow/crewai/autolog.py agent attribute capture)Medium
- Kubeflow Pipelines: frontend artifact routes skip namespace authorization, exposing other tenants' artifactsCVE-2026-118760 · Kubeflow Pipelines frontend (artifact routes, namespace authorization)Medium
- vLLM: unbounded frame count in video/jpeg base64 data URLs crashes the server with OOMCVE-2026-34755 · vLLM OpenAI-compatible API server (video/jpeg base64 multimodal path)Medium
- vLLM: no upper bound on the n parameter lets a single request OOM the API serverCVE-2026-34756 · vLLM OpenAI-compatible API server (ChatCompletionRequest/CompletionRequest n parameter)Medium
- vLLM (revision pinning): Revision pinning does not apply to all model artifactsCVE-2026-47155 · vLLM (revision pinning)Medium
This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.