Skip to content
Heist/ feed
Back to feed
High

vllm before 0.29.0 Denial of Service via Decoder Prompt Length Bypass

Type: advisorySources: OSV

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.

Products
Not specified
Affected versions
Not specified
Fixed versions
Not specified
CVEs
CVE-2026-100651
Confirmed fixed CVEs
Not specified
Updated in this feed
3 Nov 2026, 17:58 UTC · Revision 1

Source evidence

vllm before 0.29.0 Denial of Service via Decoder Prompt Length Bypass

core-osv · Original source

Published 26 Sept 2026, 13:23 UTC

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.

osv:CVE-2026-100651
Complete source record
{
  "id": "e256db7c2c73b4a353360f064f5b32078aaa8fa42f0d7c9cbc8e519267c33623",
  "url": "https://osv.dev/vulnerability/CVE-2026-100651",
  "method": "structured",
  "payload": {
    "kind": "advisory",
    "title": "vllm before 0.29.0 Denial of Service via Decoder Prompt Length Bypass",
    "summary": "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.",
    "evidence": "osv:CVE-2026-100651",
    "products": [],
    "severity": "high",
    "structured": {
      "id": "CVE-2026-100651",
      "aliases": [
        "CVE-2026-100651",
        "GHSA-3mqx-f33v-vgp9"
      ],
      "metrics": [
        {
          "type": "CVSS_V4",
          "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N"
        }
      ],
      "affected": [
        {
          "ranges": [
            {
              "repo": "https://github.com/vllm-project/vllm",
              "type": "GIT",
              "events": [
                {
                  "introduced": "0"
                },
                {
                  "fixed": "98dff2a81d747d1dba01a47f939f48c3526d4206"
                }
              ],
              "database_specific": {
                "extracted_events": [
                  {
                    "range": [
                      {
                        "introduced": "0"
                      },
                      {
                        "fixed": "0.29.0"
                      }
                    ],
                    "source": "AFFECTED_FIELD",
                    "strategy": "StandardRange"
                  },
                  {
                    "cpe": "cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*",
                    "range": [
                      {
                        "introduced": "0"
                      },
                      {
                        "fixed": "0.29.0"
                      }
                    ],
                    "source": "CPE_RANGE"
                  }
                ]
              }
            }
          ],
          "versions": [
            "v0.29.0rc6",
            "v0.29.0rc5",
            "v0.29.0rc4",
            "v0.29.0rc3",
            "v0.29.0rc2",
            "v0.29.0rc1",
            "v0.28.1rc0",
            "v0.27.2rc0",
            "v0.26.1rc0",
            "v0.25.0",
            "v0.23.1rc0",
            "v0.22.1rc0",
            "v0.21.1rc0",
            "v0.20.2rc0",
            "v0.20.1rc0",
            "v0.19.2rc0",
            "v0.19.1rc0",
            "v0.18.2rc0",
            "v0.18.1rc0",
            "v0.17.2rc0",
            "v0.18.0rc0",
            "v0.17.1rc0",
            "v0.17.0rc0",
            "v0.16.0rc1",
            "v0.15.2rc0",
            "v0.16.0rc0",
            "v0.15.0rc1",
            "v0.14.0rc1",
            "v0.14.0rc0",
            "v0.13.0rc1",
            "v0.11.1",
            "v0.11.1rc6",
            "v0.11.1rc5",
            "v0.11.1rc4",
            "v0.11.1rc3",
            "v0.11.1rc2",
            "v0.11.1rc1",
            "v0.11.1rc0",
            "v0.11.0rc1",
            "v0.10.2rc2",
            "v0.10.2rc1",
            "v0.10.1rc1",
            "v0.10.0rc2",
            "v0.10.0",
            "v0.10.0rc1",
            "v0.9.2rc1",
            "v0.9.1rc2",
            "v0.9.1",
            "v0.9.1rc1",
            "v0.9.0",
            "v0.8.4",
            "v0.8.3rc1",
            "v0.8.2",
            "v0.8.1",
            "v0.8.0rc2",
            "v0.8.0rc1",
            "v0.7.3",
            "v0.7.2",
            "v0.7.1",
            "v0.7.0",
            "v0.6.6.post1",
            "v0.6.6",
            "v0.6.5",
            "v0.6.4.post1",
            "v0.6.4",
            "v0.6.3.post1",
            "v0.6.3",
            "v0.6.2",
            "v0.6.1.post2",
            "v0.6.1.post1",
            "v0.6.1",
            "v0.6.0",
            "v0.5.5",
            "v0.5.4",
            "v0.5.3.post1",
            "v0.5.3",
            "v0.5.2",
            "v0.5.1",
            "v0.5.0.post1",
            "v0.5.0",
            "v0.4.3",
            "v0.4.2",
            "v0.4.1",
            "v0.4.0.post1",
            "v0.4.0",
            "v0.3.3",
            "v0.3.2",
            "v0.3.1",
            "v0.3.0",
            "v0.2.7",
            "v0.2.6",
            "v0.2.5",
            "v0.2.4",
            "v0.2.3",
            "v0.2.2",
            "v0.2.1",
            "v0.2.0",
            "v0.1.7",
            "v0.1.6",
            "v0.1.5",
            "v0.1.4",
            "v0.1.3",
            "v0.1.2",
            "v0.1.1",
            "v0.1.0",
            "submission"
          ],
          "database_specific": {
            "source": "https://storage.googleapis.com/cve-osv-conversion/osv-output/CVE-2026-100651.json"
          }
        }
      ],
      "modified": "2026-10-28T02:30:31.333794356Z",
      "provider": "osv",
      "references": [
        "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/100xxx/CVE-2026-100651.json",
        "https://github.com/vllm-project/vllm/security/advisories/GHSA-3mqx-f33v-vgp9",
        "https://nvd.nist.gov/vuln/detail/CVE-2026-100651",
        "https://www.vulncheck.com/advisories/vllm-before-0.29.0-denial-of-service-via-decoder-prompt-length-bypass"
      ],
      "provider_metadata": {
        "cwe_ids": [
          "CWE-400"
        ],
        "cna_assigner": "VulnCheck",
        "osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/100xxx/CVE-2026-100651.json"
      }
    },
    "related_cves": [
      {
        "value": "CVE-2026-100651",
        "evidence": "/id"
      }
    ],
    "security_relevant": true,
    "severity_evidence": "/severity/0/score"
  },
  "source_id": "core-osv",
  "document_id": 498075,
  "published_at": "2026-09-26T13:23:21.141+00:00"
}