NVIDIA Linux GPU driver: double-free of imported memory state in the kernel module

Impact

State tracking imported memory can be freed twice, corrupting the kernel allocator in a way that an unprivileged local user can steer. Memory import is how buffers move between processes and devices on a GPU node, so the path is routine traffic for frameworks rather than an exotic call. Successful exploitation gives host-level code execution from a tenant context.

Who can reach it

Local, unprivileged. A tenant container or user with access to the NVIDIA device nodes.

What to do

Apply the driver update referenced in NVIDIA bulletin 2026/5861; no fixed version is given here. Drain the node, reload or reinstall the kernel module, and reboot to be sure no stale module remains.

References

Related entries

• NVIDIA GPU driver: a user can access memory belonging to another user's processCVE-2026-47559 · NVIDIA GPU Display Driver for Windows and Linux (kernel mode layer, process memory isolation)High
• NVIDIA Linux GPU driver: unprivileged user triggers use-after-free in the kernel moduleCVE-2026-47560 · NVIDIA GPU Display Driver for Linux (kernel mode layer, object lifetime)High
• NVIDIA Linux GPU driver: ioctl input buffer size is unvalidated, giving a kernel out-of-bounds writeCVE-2026-47561 · NVIDIA GPU Display Driver for Linux (kernel mode layer, ioctl input buffer validation)High
• NVIDIA Linux GPU driver: user-triggerable NULL pointer dereference in the kernel moduleCVE-2026-47563 · NVIDIA GPU Display Driver for Linux (kernel mode layer)High
• NVIDIA Linux GPU driver: handle recycle race causes type confusion in the kernel moduleCVE-2026-47569 · NVIDIA GPU Display Driver for Linux (kernel mode layer, object handle allocator)High
• NVIDIA CUDA driver on Windows: library loaded from an uncontrolled search pathCVE-2026-47570 · NVIDIA CUDA driver for Windows (library search path)High