CVE-2026-100651

Published: Set 26, 2026 Last Modified: Set 28, 2026

ExploitDB:

Other exploit source:

Google Dorks:

HIGH 7,1

Source: [email protected]

Attack Vector: network

Attack Complexity: low

Privileges Required: low

User Interaction: none

Confidentiality: N/A

Integrity: N/A

Availability: N/A

MEDIUM 6,5

Source: [email protected]

Attack Vector: network

Attack Complexity: low

Privileges Required: low

User Interaction: none

Scope: unchanged

Confidentiality: none

Integrity: none

Availability: high

Description

AI Translation Available

vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate. When the request contains a 'features' (multimodal) payload, vllm/entrypoints/serve/disagg/serving.py builds a multimodal EngineInput directly from the caller-supplied token_ids, and GenerateRequest.token_ids (vllm/entrypoints/serve/disagg/protocol.py) is not checked against model_config.max_model_len. For multimodal processors that report skip_prompt_length_check=True (for example Nemotron Parse, Whisper, and FireRedLID), InputProcessor._validate_prompt_len() returns immediately for both encoder and decoder prompts, so an overlong prompt becomes an EngineCoreRequest and reaches the worker input-batch copy into a fixed max_model_len-wide NumPy row. A client able to reach the endpoint on an affected model configuration can therefore submit an overlong token_ids list to trigger a worker failure and denial of service. Fixed in 0.29.0.

400

Uncontrolled Resource Consumption

Draft

Abstraction: Class Structure: Simple

Common Consequences

Security Scopes Affected:

Availability Access Control Other

Potential Impacts:

Dos: Crash, Exit, Or Restart Dos: Resource Consumption (Cpu) Dos: Resource Consumption (Memory) Dos: Resource Consumption (Other) Bypass Protection Mechanism Other