Servers and applications
CVE-2026-100649
Sampler Subclass Counter Shadowing Bypasses PyNvVideoCodec Decoder Limits and GPU Memory Accounting
- Severity
- Medium 6.3
- Published
- 12 Sept 2026
- Last changed
- 6 Oct 2026
- Added to UpHeron
- 3 Nov 2026
- Fixed in
- 0.29.0
- Exploit chance
- 0.3% in the next 30 daysHigher than 28% of CVEs. EPSS by FIRST, 19 Nov 2026
- Attack
- Over the network · Only under certain conditions · No login needed · No user action needed
- Attacker can
- Slow it down
- CVSS vector
CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X- Also known as
- GHSA-j682-9xp5-rrf3
- Sources
- NVD, Project advisories (GitHub)
Description
In vLLM v0.27.0, unauthenticated video requests can select different stock sampler subclasses while using the statically configured PyNvVideoCodec backend. Because _active_decoder_slots is updated through the concrete subclass, Python creates independent shadow counters for classes such as VideoBackend, Qwen3VLVideoBackend, and Qwen2VLVideoBackend. Each class can then allocate up to hw_decoders, bypassing the intended process-wide decoder limit.
PR #49753 made PyNvVideoCodec decoder concurrency configurable, bounded the retained decoder pool, and reserved GPU memory according to hw_decoders. This bug breaks that invariant. Let A be the number of API processes sharing a GPU, C be hw_decoders, K be the number of request-selectable sampler subclasses that reach the PyNv path, and M_decoder be the retained GPU-memory cost of one decoder slot. Startup accounts for A × C slots, while runtime allocation can approach A × C × K; the unaccounted GPU-memory gap can therefore approach:
Affected software
| Software | Affected versions |
|---|---|
vLLMServers and applications · cpe:a:vllm:vllm |
|
vLLMServers and applications · github:vllm-project/vllm |
|
About vLLM
open-source software for large language model inference
- Docker Hub pulls
- 44,195,225Docker Hub, 15 Nov 2026
- GitHub stars
- 96,943GitHub, 15 Nov 2026
- Forks
- 24,836GitHub, 15 Nov 2026
- Latest release
- v0.33.0rc3GitHub, 15 Nov 2026
- Released
- 2 Nov 2026GitHub
- License
- Apache Software License 2.0Wikidata
- First published on GitHub
- 9 Feb 2023GitHub
Website Source code Docker image
Vulnerabilities in vLLM
- In our list
- 169
- Published in the last 12 months
- 150
- Critical, last 12 months
- 6
- High, last 12 months
- 63
- Known exploited
- 0
- CVE-2026-130686vLLM 0.6.0 before 0.33.0 Resource Exhaustion via Multi-Image RequestsHigh 8.7 18 Nov 2026
- CVE-2026-130684vLLM before 0.33.0 Path Traversal via speculators_config verifier.name_or_pathMedium 5.3 18 Nov 2026
- CVE-2026-130685vLLM 0.28.0 before 0.33.0 Unauthenticated DoS via Out-of-Range Prompt Token IDHigh 8.7 18 Nov 2026
- GHSA-xgh8-qwjh-q853/v1/systemone bypasses the --trust-request-chat-template gateMedium 6.5 18 Nov 2026
- CVE-2026-129356vllm-project vllm Common Prefix single_type_kv_cache_manager.py race conditionLow 2.3 15 Nov 2026
- CVE-2026-126394vLLM 0.30.0 before 0.31.0 Resource Exhaustion via OpenPangu Video Backend num_framesMedium 6.9 13 Nov 2026
References
- NVD https://nvd.nist.gov/vuln/detail/CVE-2026-100649
- CVE.org https://www.cve.org/CVERecord?id=CVE-2026-100649
- GitHub Advisory Database https://github.com/advisories/GHSA-j682-9xp5-rrf3
- github.com https://github.com/vllm-project/vllm/security/advisories/GHSA-j682-9xp5-rrf3