Evidence · DiggingBeagle record
vLLM Jinja renderer CPU exhaustion, CVE-2026-115911
Advisory on unbounded tokenizer chat templates monopolizing a shared server-renderer worker.
- Source role
- primary disclosure
Evidence record
Advisory on unbounded tokenizer chat templates monopolizing a shared server-renderer worker.
Claim-level citations (2)
- supportsvLLM unbounded Jinja chat templates could monopolize a shared renderer: The vLLM advisory describes a small unbounded Jinja chat template that stalls a shared renderer worker and blocks later requests, under an attacker-controlled template input path.
GHSA-542m-9234-8gvf: Jinja2 rendering worker, crafted template and patch
- supportsvLLM unbounded Jinja chat templates could monopolize a shared renderer: A tokenizer chat template containing an unbounded Jinja loop occupied the vLLM shared rendering worker while subsequent inference requests waited. The attacker needed to supply a model tokenizer configuration or otherwise control a reachable chat-template input path, not merely write ordinary prompt text. A small template could create an indefinitely stalled renderer or large computation and deny service to other model users.
GHSA-542m-9234-8gvf: Jinja2 rendering worker, crafted template and patch