vLLM denial of service in prompt logprobs with cached prefixes
vllm-project vllm
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| vllm-project vllm |
|
|
Original description
vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.
From the CVE record · VulnCheck
References
- github.com/vllm-project/vllm/security/advisories/GHSA-rhcx-5729-88vgVendor advisory
- vulncheck.com/advisories/vllm-from-0.8.5-before-0.31.0-cross-tenant-dos-via-uninitialized-logprobstensorsAdvisory
- github.com/CVEProject/cvelistV5/tree/main/cves/2026/116xxx/CVE-2026-116180.json
- nvd.nist.gov/vuln/detail/CVE-2026-116180