Database/AI/ML frameworks & serving
llama.cpp: reachable assertion in the Mirostat sampler aborts the server remotely
Impact
A remote request can drive the Mirostat sampler into a reachable assertion, aborting the process. Where llama.cpp is exposed as a server, that is an unauthenticated kill of the serving replica, and because the model has to be reloaded into GPU memory on restart, recovery is a cold start affecting everyone on that replica rather than a single failed request. The record is thin: it comes from a third-party database, names only the sampler and the file, and gives no reproduction detail or exploit mechanism, and the only impact claimed is low availability.
Who can reach it
Remote and unauthenticated per the record - a request to a llama.cpp server that reaches Mirostat sampling. The record does not say which parameter values are required.
What to do
No upstream fix as of this record: the project was notified through pull request 29992 and had not acted. Mitigate by disabling Mirostat sampling where you can pin sampler settings server-side, rejecting client-supplied Mirostat parameters at the gateway, and keeping a supervisor in place to restart the process. Track the pull request for a fix; no fixed version exists to name.
References
Related entries
- llama.cpp: integer overflow in GGUF tensor shape validation leads to out-of-bounds memory accessCVE-2026-116829 · llama.cpp gguf_init_from_reader (tensor element-count validation)Medium
- llama.cpp: integer overflow in GGUF tensor size padding yields an undersized data blob and out-of-bounds readsCVE-2026-116840 · llama.cpp (gguf_init_from_reader tensor size padding)Medium
- Crawl4AI: CacheValidator freshness check bypasses the egress proxy, giving blind SSRF from the crawler containerCVE-2026-119122 · Crawl4AI Docker server (CacheValidator freshness check)Medium
- Hugging Face Accelerate: unsanitized shard paths in a checkpoint index give arbitrary file read and hangsCVE-2026-69112 · Hugging Face Accelerate (load_checkpoint_in_model / load_checkpoint_and_dispatch weight_map handling)Medium
- stable-diffusion.cpp: safetensors loader trusts data_offsets, reading process memory past the tensor data sectionCVE-2026-70660 · stable-diffusion.cpp safetensors loader (src/model_io/safetensors_io.cpp)Medium
- llama.cpp ggml RPC server: unvalidated tensor op and op_params in deserialize_tensorCVE-2026-78147 · llama.cpp ggml RPC server (deserialize_tensor op / op_params validation)Medium
This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.