CVE-2026-118375 PUBLISHED

llama.cpp b8841 before b11865 contains an out-of-bounds write vulnerability in the ggml-rpc RDMA transport rdma_recv() function that allows unauthenticated peers to overflow the rpc-server stack. Attackers advertising RDMA capabilities in the HELLO message can send up to 256 KiB into a 1-byte stack variable in rpc_serve_client(), crashing the server or potentially executing code.

• Sohaib Harraoui (Ostorlab) finder
• Sohaib Harraoui (Ostorlab) remediation developer

• https://github.com/ggml-org/llama.cpp/pull/30399
• https://github.com/ggml-org/llama.cpp/commit/9a0fde9967fa1a35b6e5dd1f47e1a3648b6fecc1
• https://github.com/ggml-org/llama.cpp/blob/620d2225cc436ab0b001e3200e590e3b286ef9ab/ggml/src/ggml-rpc/transport.cpp#L510-L528
• llama.cpp b11865 Release Notes
• https://github.com/ggml-org/llama.cpp
• VulnCheck Advisory: llama.cpp b8841 before b11865 Stack-based Buffer Overflow via ggml-rpc RDMA rdma_recv()

• Out-of-bounds Write CWE