Update for proprietary NVIDIA drivers 440.100 and 390.138 to fix vulnerabilities

NVIDIA Company presented new proprietary driver releases NVIDIA 440.100 (LTS) and 390.138 ("legacy" for GPU GF1xx "Fermi"), which eliminated dangerous vulnerabilities, potentially allowing you to elevate your privileges in the system. The driver is available for Linux (ARM, x86_64), FreeBSD (x86_64) and Solaris (x86_64).

• CVE‑2020‑5963 is a vulnerability in the CUDA Driver Interprocess Communication API that could lead to a denial of service, elevated code execution, or information leakage.
• CVE‑2020‑5967 is a race condition vulnerability in the UVM driver that could lead to a denial of service.

Version 440.100 also supports new GPUs
GeForce GTX 1650 Ti, GeForce GTX 1650 Ti with Max-Q, GeForce RTX 2060 with Max-Q and Quadro T1000 with Max-Q, anonymized device alias "Connector-N" added for X11 settings, which can be used in the ConnectedMonitor option for emulation connecting a monitor without knowing the available connection methods.
Version 390.132 added compatibility with the Linux 5.6 kernel and Oracle Linux 7.7, and added support for PRIME Synchronization for Linux 5.4 kernel systems.

Additionally, started testing a beta version of the new branch 450.x, in which the following improvements can be noted:

• Added support for GPU A100-PCIE-40GB, A100-PG509-200,
  A100-SXM4-40GB, GeForce GTX 1650 Ti, GeForce RTX 2060 with Max-Q and
  Quadro T1000 with Max-Q;

• The Vulkan API now supports direct display on displays connected via DisplayPort Multi-Stream Transport (DP-MST);
• Added support for OpenGL extension glNamedBufferPageCommitmentARB;
• Added libnvidia-ngx.so library with implementation of technology support NVIDIA NGX;
• Improved detection of Vulkan-enabled devices on systems with an X.Org server;
• The libnvidia-fatbinaryloader.so library was removed from the distribution, the functionality of which was distributed among other libraries;
• The means of dynamic power management are expanded by the ability to turn off the power of the video memory;
• VDPAU adds support for 16-bit video surfaces and the ability to accelerate decoding of HEVC 10/12-bit streams;
• Added support for "Image Sharpening" mode for OpenGL and Vulkan applications;
• Removed IgnoreDisplayDevices X server configuration option;
• Added support PRIME Synchronization to render through another GPU on the system using the x86-video-amdgpu driver. Provided the ability to use screens connected to the NVIDIA GPU as "Reverse PRIME" to display the results of another GPU in systems with multiple GPUs;
• Threaded optimizations for OpenGL are disabled by default due to regressions in some situations.

Source: opennet.ru