What can I help you with?
NVIDIA Messaging Accelerator (VMA) Documentation Rev 9.8.71

Changes and New Features

Feature/Category

Description

VMA User Memory Allocation

Added support for custom memory allocators for VMA’s RX/TX buffers, giving applications more control and improving Zero-Copy performance. For more information, see VMA User Memory Allocation.

  • Multi Packet Receive Queue beta functionality is removed as of VMA v9.3.1

  • RDMA experimental verbs library (mlnx_lib)

  • VMA v9.3.1 and up do not enforce the disable_raw_qp_enforcement option; use the CAP_NET_RAW option instead

Important Notes

Note

We recommend using libnl3 as it is the latest version and includes fixes related to libnl1

Note

Bonding Active-Backup (mode 1) is supported with limitations shown in .Known Issues v9.4.0 section.


© Copyright 2025, NVIDIA. Last updated on Jun 8, 2025.