Details
Option 2 composes with #51969 and needs no per-backend audit; I would favour it, but option 1 is the more conservative change and matches what has already been merged.
Affected versions
>= 0.24.0, through v0.29.1rc0 and main @ b23433088.
Lower bound established by probing release tags through the GitHub contents API; no version below is inferred, each was read out of the file at that tag.
Confirmed present — Qwen samplers reading unclamped `max_frames` (max_frames = kwargs.get("max_frames", 768) inside Qwen2VLVideoBackend / Qwen3VLVideoBackend in vllm/multimodal/video.py):
| ref | Qwen3VLVideoBackend present | unclamped max_frames |
|---|---|---|
| v0.23.0 | no (class does not exist) | n/a |
| v0.24.0 | yes | yes |
| v0.25.0, v0.26.0, v0.27.0, v0.28.0, v0.29.0, v0.29.1rc0 | yes | yes |
| main @ b23433088 | yes | yes |
Confirmed present — request-level `media_io_kwargs` (field on the chat request model, and VideoMediaIO.merge_kwargs present): every ref probed, v0.19.0 through v0.29.1rc0 and main @ b23433088.
Confirmed absent — any `num_frames` ceiling clamp (PR #51969 unmerged): every ref probed, v0.19.0 through v0.29.1rc0 and main @ b23433088.
Not resolved, and why:
- The introducing commit/PR for
Qwen2VLVideoBackend/Qwen3VLVideoBackend. My clone is shallow (--depth=300), sogit log -S'max_frames'cannot reach it; the v0.23.0 → v0.24.0 boundary above is the tightest bound I established by probing release tags. - The rc tags between v0.23.0 and v0.24.0, to tighten the lower bound to a specific release candidate.
- Whether a
max_framesknob on a *differently named* pre-v0.24.0 backend is separately affected — at v0.23.0GLMGAVideoBackendalready carriedmax_frames = kwargs.get("max_frames", 640)with no clamp, and that clamp was only added on 2026-09-04 by8b6de0eb9. Versions between are plausibly affected through GLMGA rather than Qwen; I did not test that path. - Whether GHSA-vxqj-p4gw-9h4c's own affected range differs, which I cannot see — the advisory returns 404 to me.
No dependency versions are asserted anywhere in this report.
Weaknesses of this report, stated upfront
- No GPU was used. This host has no CUDA device, so the HTTP results come from the in-tree GPU-less render server rather than a full
vllm serve. That server runs the real frontend — the same request model, the samemerge_media_io_kwargs→VideoMediaIO.merge_kwargs→MediaConnector.fetch_videochain, the same/tokenizeroute — and media decoding happens entirely in the frontend, so I do not believe the engine's presence changes the result. I have not confirmed that on a GPU deployment, and a reviewer may reasonably want that repeated undervllm serve. - The in-process measurements (the #51969 comparison table) were taken by importing the tree directly at
b23433088, verified by__file__, with novllmwheel installed. - Amplification figures are a floor. The OpenCV build available here offers only
mp4v/XVID, giving ~2 200× compression on static content. An attacker using x264/x265 would do materially better for the same frame count. - The two large figures in the Impact section (9.73 GiB RSS; the OOM kill) were measured on the
num_framespath, not this one, and are labelled as such. - Per-frame pixels remain bounded by
VLLM_MAX_IMAGE_PIXELS; this concerns the unbounded frame *count*, which multiplies it. VLLM_MAX_MEDIA_DOWNLOAD_SIZE_MB(default 256) caps the compressed size of an HTTP-fetched video but not adata:URI, whichMediaConnector.load_from_urldispatches before any size logic is reached.
---
*Prepared with AI assistance (Claude), per the repository's contributing guidance on disclosing AI-assisted contributions. All findings were verified by executing vLLM's own code at the commits cited; the analysis and the claims are my own.*
*Reported by Eva Crystal / 0xiviel (XSource Security).*