Decode-worker crash via unvalidated tensor-parallel size
Published Sep 21, 2026 · Updated Sep 21, 2026
Uncontrolled resource consumption in vLLM through 0.29.0 allows attackers to crash decode workers via crafted completion requests. The completion handler accepts caller-controlled kv_transfer_params.tp_size and uses it to size NIXL decode-side state without enforcing a bound. The request must reach a decode worker in a prefill/decode disaggregated deployment; exhaustion can invoke the kernel OOM killer and terminate that worker.
Summary
What happened
Uncontrolled resource consumption in vLLM through 0.29.0 allows attackers to crash decode workers via crafted completion requests. The completion handler accepts caller-controlled kv_transfer_params.tp_size and uses it to size NIXL decode-side state without enforcing a bound. The request must reach a decode worker in a prefill/decode disaggregated deployment; exhaustion can invoke the kernel OOM killer and terminate that worker.
The record
- CVE
- CVE-2026-94626
- Published
- Sep 21, 2026
- Updated
- Sep 21, 2026
- Vendor
- vLLM Project
- Product
- vLLM
- Classifications
- T1499
- Attack vector
- network
- Privileges
- Unavailable
Timeline
How it unfolded
- Sep 21, 2026CVE publishedPublication date reported by the CVE source.
- Sep 21, 2026Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=0 <=0.29.0
What conditions does exploitation require?
What is affected?
Published CVSS scores
No CVSS assessment is available in this record.
CVSS describes severity. EPSS estimates exploitation probability.
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Activity data is delayed. The latest report is Sep 28, 2026.
Weakness, pattern, technique
Public exploit references
No public exploit references are available in this record.
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Get started to see how Hinoki identifies affected software and validates exploitability in your environment.
Get started