vLLM: VLLM through 0.29.0 contains a denial of service vulnerability in the NIXL connector's prefix…
vLLM
More for vLLM RSS feed for vLLM
- Severity
- High, CVSS 7.5
- CVSS vector
- CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
- Record
- CVE-2026-94623
- Affects
- vLLM
- Known exploited
- Not listed in the CISA Known Exploited Vulnerabilities catalog
- Published
- Record first published by the source
- September 21, 2026
- Source last modified
- September 29, 2026
Description
vLLM through 0.29.0 contains a denial of service vulnerability in the NIXL connector's prefix caching implementation that fails to properly validate block counts across multi-prompt completion requests in prefill/decode disaggregated deployments. Attackers can trigger an assertion failure in NixlBaseConnectorWorker._apply_prefix_caching by submitting completion requests with multiple prompts of varying lengths, causing the decode worker to terminate and become unavailable until restarted.
Description quoted from the NVD record.
Who is affected
This flaw affects vLLM, a program that runs AI language models so other software can send them questions and get answers. It affects vLLM versions through 0.29.0 that are set up to split the work of answering requests between separate servers and that use the NIXL connector, a part that passes reused work between those servers. Someone could send requests that each contain several prompts of different lengths. This makes the server that finishes the answers shut down, and it stays unavailable until it is restarted. This flaw is rated High.
What to do
- The person who installed or runs vLLM for your organization should check whether you use version 0.29.0 or earlier with the work split between separate servers through the NIXL connector.
- No fixed version is named for this flaw yet, so check the vendor advisory linked on this page for a fixed release and install it once one is available.
- If an IT provider or vendor runs vLLM for your organization, send them this advisory and ask when the update will be done.
- Until then, if the server that finishes the answers stops working, restart it to bring it back.
Analysis written from the source record as of September 29, 2026 and reviewed before publishing.