Changes and New Features

NVIDIA Accelerated IO (XLIO) Documentation Rev 3.20.7

Feature/Category

Description

SocketXtreme API

• Renew support for SocketXtreme API. See section 9.4

• Added support for GRO

• Added support for TX polling

Packet Filtering

Renew support for Packet Filtering. See section 9.5

TCP Acceleration (kernel bypass) for NGINX

• Added support for NGINX proxy mode.

• Renamed XLIO_NGINX_DISTRIBUTE_CQ to XLIO_DISTRIBUTE_CQ

Memory Allocation

• Added support for 64K page Kernel

• Added support for allocation of different Hugepage sizes

• Added memory utilization report

• Added XLIO_MEMORY_LIMIT/ XLIO_MEMORY_LIMIT_USER/ XLIO_HUGEPAGE_LOG2/ XLIO_PRINT_REPORT paramters.

• Removed ‘CONTIG’ option from XLIO_MEM_ALLOC_TYPE.

Ring Migration

Changed default of XLIO_RING_MIGRATION_RATIO_RX to disabled.

Socket options

Added support for IP_BIND_ADDRESS_NO_PORT socket option.

Statistics

Added counters for destructed sockets. Added CSV output.

Internal Thread

Added ‘delegate’ option to XLIO_TCP_CTL_THREAD.

Bug Fixes

See Bug Fixes section.

Warning

As of XLIO v3.10, DEBUG-level logs will not be available in the release builds. Alternatively, you can use libxlio-debug.so to obtain those logs. For details on how to use libxlio-debug.so, please refer to Libxlio-debug.so section.

Warning

libnl1 is no longer supported. Please use libnl3 instead.

Warning

Bonding Active-Backup is not supported.
RoCE LAG is a feature meant for mimicking Ethernet bonding for IB devices and is available for dual-port cards only. XLIO cannot offload traffic in cases where RoCE LAG is enabled too. In RoCE LAG mode, instead of having an IB device per physical port (for example, mlx5_0 and mlx5_1), only one IB device is present for both ports.

Warning

Direct Packet Control Plane (DPCP) provides a unified, flexible interface for programming NVIDIA NICs and comes as part of OFED. The DPCP version must be v1.1.36 and above. For further details, please see Direct Packet Control Plane (DPCP).

Warning

TLS Rx offload related notes:

  • TLS Rx offload supports up to 64K concurrent connections

  • TLS Rx offload for IPv6 is not supported as it may cause unexpected behavior

  • TLS Rx offload requires OpenSSL 3.0.2 or higher and kTLS support from the kernel

© Copyright 2023, NVIDIA. Last updated on Nov 3, 2023.