CVE-2026-90878
Received Received - Intake

Resource Exhaustion in vLLM via Jinja Template Rendering

Publication date: 2026-09-15

Last updated on: 2026-09-15

Assigner: VulDB

Description
A vulnerability was determined in vllm-project vLLM up to 0.27.1. This affects an unknown part of the file /v1/chat/completions of the component Jinja Template Rendering. This manipulation of the argument chat_template causes resource consumption. The attack can be initiated remotely. The exploit has been publicly disclosed and may be utilized. The pull request to fix this issue awaits acceptance.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published
2026-09-15
Last Modified
2026-09-15
Generated
2026-09-15
AI Q&A
2026-09-15
EPSS Evaluated
N/A
NVD
CVE-2026-90878
EUVD
EUVD-2026-78268

Affected Vendors & Products

Showing 2 associated CPEs
Vendor Product Version / Range
vllm-project vllm to 0.27.1 (inc)
vllm-project vllm 0.27.1

Helpful Resources

Exploitability

CWE
CWE Icon
KEV
KEV Icon
CWE ID Description
CWE-404 The product does not release or incorrectly releases a resource before it is made available for re-use.
CWE-400 The product does not properly control the allocation and maintenance of a limited resource.

Attack-Flow Graph

AI Quick Actions

Instant insights powered by AI
Executive Summary

This vulnerability is a denial-of-service issue in vLLM up to version 0.27.1. It involves the Rust frontend rendering Jinja chat templates from request bodies without resource limits. Attackers can craft small templates that cause excessive CPU consumption due to O(N^depth) complexity, monopolizing worker threads and halting chat completions.

Detection Guidance

Monitor for unusually high CPU usage on vLLM workers during chat completion requests. Check for repeated requests containing large or deeply nested Jinja templates in the /v1/chat/completions endpoint. Use system monitoring tools like top, htop, or ps to observe worker thread consumption.

Impact Analysis

The impact includes service disruption where legitimate users cannot access chat completions due to all worker threads being occupied by malicious template rendering. It can lead to degraded performance or complete unavailability of the vLLM service, even without GPU or model weights.

Compliance Impact

This vulnerability could lead to denial-of-service conditions by consuming excessive CPU resources through crafted Jinja templates, potentially disrupting services handling sensitive data. For GDPR, this may impact availability of personal data processing systems, while for HIPAA, it could interrupt healthcare services processing protected health information. The lack of resource limits in the Rust frontend exacerbates these risks.

Mitigation Strategies

Reject all chat_template parameters in /v1/chat/completions requests at the network or application level. Update vLLM to the latest version once the fix is merged. Temporarily limit concurrent requests to reduce impact if exploitation is suspected.

Chat Assistant

Ask questions about this CVE
Hi! I’m here to help you understand CVE-2026-90878. Ask me anything about the vulnerability, its impact, or mitigation strategies.
0/70

EPSS Chart