NVIDIA 900-9D3B6-00SV-A DPU Firmware 32.43.1014

Firmware which is added at the time of manufacturing, is used to run user programs on the device and can be thought of as the software that allows hardware to run. Embedded firmware is used to control the functions of various hardware devices and systems, much like a computer’s operating system (OS) controls the function of software applications. Firmware may be written into read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory.

NVIDIA BlueField-3 DPU provides innovative acceleration, security, and efficiency in every host. BlueField-3 data center infrastructure combines the power of the NVIDIA ConnectX-6 Dx with programmable Arm cores and hardware offloads for software-defined storage, networking, security, and management workloads.

NVIDIA BlueField-3 also delivers superior performance, security, and reduced TCO for cloud computing platforms, enabling organizations to efficiently build and operate virtualized, containerized, and bare-metal infrastructures at massive scale.

Changes and New Features:

- Migrated PCC NP solution from ACE hardware platform to DPA hardware platform. The new capability is applicable to the following 2 modes: PCC_INT_EN=True and PCC_INT_NP_RTT_DATA_MODE=INGRESS_BYTE / PCC_INT_EN=True and PCC_INT_NP_RTT_DATA_MODE=NO_DATA. The first mode is used to support ZTRCC RX bytes in RTT response.
- Added support for HPCC2 custom header insertion in RTT request packets for DOCA PCC. The capability will be supported when setting ROCE_CC_STEERING_EXT = ENABLED.
- Added support for a second emulation VirtIO blk and net device on the same vHCA to enable switching to the second emulation device and reduce downtime.
- Added the option to indicate an error CQE event on every selected function per eSwitch manager. This indication is defined as a new WQE including the relevant information about the error (such as: syndrome, function_id, timestamp, QPs num etc.).
- Extended kernel lockdown permission set. The following sub-operations can now be called by tools (permission TOOLS_RESORCES) using new HCA capability bitmask field: tool_partial_cap.

The 5 sub-operations are:
- QUERY_HCA_CAP with other function
- QUERY_VUID with direct data
- QUERY_ROCE_ADDRESS with other vport
- SET_HCA_CAP with other function
- POSTPONE_CONNECTED_QP_TIMEOUT with other vport

The new added caps are:
- tool_partial_cap.postpone_conn_qp_timeout_other_vport,
- tool_partial_cap.set_hca_cap_other_func
- tool_partial_cap.query_roce_addr_other_vport
- tool_partial_cap.query_vuid_direct_data
- tool_partial_cap.query_hca_cap_other_func

- Added support for QoS scheduling across multiple E-Switches grouped in a LAG. VPort members of a Physical Function can be added to a rate group from another Physical Function and rate limits of the group will apply to those VPort members as well.
- Added 'table_type_valid' and 'table_type' fields to the steering action (STC) "Jump To Flow" table parameters to enable the user to jump from NIC_TX to FDB_TX and bypass the ACL table.
- Enabled hop reduction by bypassing NIC domain in various use cases. Such action r educes the number of hops (improves PPS) to deal with mass number of flows and devices. To enable this new capability, a new STC action type "JUMP_TO_FDB_RX" was added to allow jumping into the RX side of a table.
- Added a new scheduling element type ("TC_ARB") capability in the VQoS domain (Virtual Quality of Service), to support TC arbitration between functions (VPORTs).
- Enabled hotplug/hotunplug during device's power off or power cycle to prevent the device from getting stuck.
- Added support for 2-steps-hotplug capability. The device is plugged with "free" status by default, and it will not appear on the bus until being modified to "hotplug" status.

Bug Fixes:

- Added a recovery step in case of CQ doorbell getting lost during VF migration.
- Fixed an issue that occasionally caused PTP accuracy degradation for port speed configured to 1G or 10G.
- Modified the UEFI driver MFLR send behavior upon boot. Now MFLR is send only if there is a setting change and not per boot.
- Fixed the query for FACTORY default NV configuration values. The firmware always returned the "next" value to be applied.
- Changed the PCI Gen4/5 default CTLE VGA gain.
- Changed the static configuration for Gen4/5 in the SerDes to support high gain signal.
- Some pre-OS environments may fail when sensing a hot-plug operation during their boot stage.
- Added address validation in MLNX OEM CMD 0x0032 (get debug info) to be 4-bytes aligned.
- Fixed an issue that resulted in MCTP SMBUS TX buffer corruption in multi-bmc topology. To resolve it, a separate buffer for each BMC was implemented.

Known Issues:

- NC-SI pass-through requires the user to allocate a MAC address to the platform BMC.
- NVIA register is not allowed for external host if any field of EXTERNAL_HOST_PRIV or EXTERNAL_HOST_PRIV_FAST TLVs is not set as the default.
- When configuring BlueField-3 Arm cores as PCIe root-complex, all non-mlx5 devices must always set the BlueField-3’s IOMMU to disabled or passthrough mode. Turning IOMMU “ON” requires special handling of interrupts in the driver or the use of polling.
- The supported DDR5 link speed in SKU B3220, is 5200 MT/s.
- SW_RESET with a pending image is currently not supported.
Occasionally, the device may hang when there a hot plug is performed from a unknown direction.
- The I2C clock fall time is lower than the 12ns minimum defined in the I2C-bus specification.
- When connecting to a High Speed Traffic Generator in 400G speed, the linkup time may takes up to 3 minutes.
- External flash access such as flash read using the MFT tools will fail if there is a pending image on the flash.
- On BlueField-3 devices, from DOCA 2.2.0 to 32.37.1306 (or lower), the host crashes when executing partial Arm reset (e.g., Arm reboot; BFB push; mlxfwreset).
- When unloading the network drivers on an external host, sync1 reset may be still reported as 'supported' although it is not. Thus, initiating the reset flow may result in reset failure after a few minutes.
- When connecting to a Spirent switch in 400G speed, the linkup time may takes up to 3 minutes.
- PCIe PML1 is disabled.
- Unexpected system behavior might be observed if the driver is loaded while reset is in progress.
- When performing PCIe link secondary-bus-reset, disable/enable or mlxfwreset on AMD based Genoa systems, the device takes longer then expected to link up, due to a PCIe receiver termination misconfiguration.
- he firmware rollback fails for the signature retransmit flow if the QPN field is configured in the mkey (as it only allows the given QP to use this Mkey) as the firmware rollback flow relies on an internal QP that uses the mkey.
- Socket-Direct is currently not supported.

About Network Adapter Drivers:

When connected, the operating system usually installs a generic driver that helps the computer to recognize the newly attached device.

However, proper software must be applied if you want to make use of all features that the network adapter has available. This task also allows computers to properly recognize all device characteristics such as manufacturer, chipset, technology, and others.

Updating the adapter's drivers and utilities version might improve overall performance and stability, increase transfer speeds, fix different compatibility problems and several network-related errors, as well as bring various other changes.

To install this release, simply get the package, extract it if necessary, run the setup, and follow the instructions displayed on-screen. When done, don't forget to perform a system restart and reconnect the network adapter to make sure that all changes take effect properly.

Without further ado, if you intend to apply this version, click the download button and install the package. Moreover, check with our website as often as possible so that you don't miss a single new release.