vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mm_audio.py and phi4mm_utils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forward_attention to materialize and retain full attention matrices in self.attn. The scaled dot-product attention branch does not materialize these matrices. The resulting quadratic GPU-memory use can cause an out-of-memory condition and terminate EngineCore for all users. This issue is fixed in version 0.32.0.
References
| Link | Resource |
|---|---|
| https://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc89928 | Patch |
| https://github.com/vllm-project/vllm/pull/60614 | Patch |
| https://github.com/vllm-project/vllm/releases/tag/v0.32.0 | Release Notes |
| https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm | Exploit Patch Vendor Advisory |
Configurations
History
28 Oct 2026, 15:38
| Type | Values Removed | Values Added |
|---|---|---|
| References | () https://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc89928 - Patch | |
| References | () https://github.com/vllm-project/vllm/pull/60614 - Patch | |
| References | () https://github.com/vllm-project/vllm/releases/tag/v0.32.0 - Release Notes | |
| References | () https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm - Exploit, Patch, Vendor Advisory | |
| CPE | cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:* | |
| First Time |
Vllm vllm
Vllm |
26 Oct 2026, 17:17
| Type | Values Removed | Values Added |
|---|---|---|
| New CVE |
Information
Published : 2026-10-26 17:17
Updated : 2026-10-28 15:38
NVD link : CVE-2026-115913
Mitre link : CVE-2026-115913
CVE.ORG link : CVE-2026-115913
JSON object : View
Products Affected
vllm
- vllm
CWE
CWE-770
Allocation of Resources Without Limits or Throttling