1.2.1.1.1 General

• None in this release.

1.2.1.1.2 Firmware/NVM/NVM Update

• None in this release.

1.2.1.1.3 Linux

• None in this release.

1.2.1.1.4 FreeBSD Driver

• None in this release.

1.2.1.1.5 Windows Driver

• Enhancements to completion path correctness and stability, doorbell behavior adjustments, queue/ring concurrency hardening, better error propagation mechanisms. [15018815760]

1.2.1.1.6 Windows Server

• None for this release.

1.2.1.1.7 ESX Driver

• None for this release.

1.2.1.1.8 Pre-Boot

• Added SMBIOS Type 9 SegmentGroupNum to check that the logic matches the SMBIOS structure with the NIC instance. [15019458988]

1.2.1.2.1 General

• None for this release.

1.2.1.2.2 Firmware/NVM/NVM Update

• None for this release.

1.2.1.2.3 Linux Driver

• None for this release.

1.2.1.3.1 Windows Driver

• None for this release.

1.2.1.3.2 Firmware/NVM/NVM Update

• Module FINISAR CORP.FTRJ8519P1BNL is unable to regain link after issuing a link reset. Recovery requires removing the module from both E825 and switch to reacquire link. [22021219549]

1.2.1.4.1 General

• In certain situations, the module temperature read fails to return a valid value, triggering a module error message. The new temperature implementation is fully reworked to function asynchronously from the request. [15019523533]

1.2.1.4.2 Linux Driver

• User may observe results in WARN_​ON call trace and “unable to move VSI idx 1 into aggregator 1 node”, when using the devlink reload command in ICE Linux driver. [22020980371]

1.2.1.4.3 Firmware/NVM/NVM Update

• None for this release.

1.2.1.4.4 FreeBSD Driver

• Due to an error in the FreeBSD ice-2.1.8 component archive build process, it lacks the latest version of the Dynamic Device Personalization package. To avoid entering Safe Mode and limiting adapter functionality, use the DDP package included in the FreeBSD OS. [21044824146]

1.2.1.4.5 ESX Driver

• None in this release.

1.2.1.4.6 Windows Driver

• None in this release.

1.2.1.4.7 Pre-Boot

• Added SMBIOS Type 9 SegmentGroupNum to check that the logic matches the SMBIOS structure with the NIC instance. [15019458988]

1.3.1.1.1 General

• Windows Server 21H1 OS is unable to save memory dump (crash dump) on disk. It is considered to be an OS defect. [18014988594]

• DPDK traffic is stopped after FLR reset. This issue has been documented in the rte_​eth_​dev_​reset API. [16018959537]

  Workaround: testpmd can be used to recover a VF after a reset.

• When a VF reset happens, testpmd will print out “port reset” event to the console.

• Use the “port reset” command to call rte_​eth_​dev_​reset, and everything will go back to normal

1.3.1.1.2 Firmware/NVM/NVM Update

• The firmware aligns to DMTF spec clarification. The port schema defines that all the LLDPRecieve sub-properties shall not be present in the resource if the property was not received. [15019109850]

1.3.1.2.1 Linux

• East-West (VF-VF) traffic fails when using switchdev SR-IOV link aggregation (LAG) in active-backup mode [16029737566]

• Configuration of a link down on PF0 in LAG mode resulted in a rping test failure. [15018677889]

• The system crash is observed when user is changing ring size (ethtool -G) when Jumbo frames are enabled (MTU 9000) with traffic present. [22021169950]

• When user sets more than 8 VLANs for trusted VF, and then moves VF as untrusted, the VLAN configuration will be lost. [18032096535]

  Workaround: To avoid losing VLAN configuration, user shall first reduce VLANs configuration allowed for untrusted VF (not more than 8 VLANs per VF), and then switch the VF to untrusted mode.

1.3.1.2.2 FreeBSD Driver

• On a Linux host, iavf interfaces in FreeBSD-13.0 guests may experience poor receive performance during stress. [18019784282]

  Workaround: The workaround for this behavior is to do the power cycle of the setup to see the assigned DCB-MAP is reflecting.

1.3.1.2.3 RDMA Driver

• None for this release.

1.3.1.2.4 VMware Driver

• None for this release.

1.3.1.2.5 Windows Driver

• None for this release.

1.3.1.2.6 ESX Driver

• VF-to-PF and PF-to-VF traffic may fail intermittently when SR-IOV Trusted Mode is used with ENS on ESXi. [18040914960]

  Workaround: Avoid ENS + SR-IOV Trusted Mode combinations in production/cert flows; use a supported networking mode until VMware support is available (targeted with newer VCF stack).

• Running Unreliable Datagram (UD) RDMA mixed traffic with more than 2 QPs may lead to a receiver side UD application hang. [18015677916]

  Workaround: Restart the RDMA UD application. This is not expected to impact storage (NVMeoF, iSER, VSAN) applications since they do not rely on UD communication.

• <VMWARE> When instantiating the maximum number of VFs in NSX-T, adding a Transport Node afterwards might fail due to timeout. [18021236660]

• DDP updates triggered from ENS uplink ports on mixed mode devices may take longer due to extended driver recovery, and the workaround is to wait a minimum of 30 seconds before issuing any subsequent DDP commands. [16029636950]

1.3.1.3.1 General

• None for this release.

1.3.1.3.2 Firmware/NVM/NVM Update

• After advertising the link to 1G in the MRVL system, the link down and up happened 2 times in SUT and a single time in TG. [16022825930]

• Using the EPCT tool to change the port configuration requires 2 reboots to complete the programming process. Blank mode or PTP initialization failures may be observed after a single reboot and will be resolved after performing a second reboot. [16021278765]

• When loading the ICE driver with cages populated, the system displays the error message “Module is not present.” [1509371756]

• When 100 MB option is selected in Windows Device Manager, a link cannot be established. [18017637026]

• Lane Reversal is broken preventing proper functionality of 2x1x50g port option on quad 1. CPI opcode 0x67 PortLaneOrder does not support setting lane 1 as the autoneg lane and returns error code 1 (Configuration Error). [22016503876]

1.3.1.3.3 Linux Driver

• ethtool gives an invalid argument error when using ethtool -m <interface>. [22021254023]

• Tx Clock Source Selection (NVM) - Feature Not Supported. Tx Clock Source selection is not fully implemented and validated in this release. Configuration of the Tx Clock Source through software interfaces is not supported. To ensure correct device operation, the Tx Clock Source field (Port Options 1, bits [15:14]) must remain set to the default value of 0x3 (“Do Not Change”). Customers should not modify the Tx Clock Source setting and should not rely on non-default values. Any software or management tool capable of modifying this field should not be used to select an alternate Tx Clock Source. Using values other than 0x3 may result in undefined behavior, including link bring-up failures, PHY communication errors, or system instability. [15019916102]

Workaround: Keep Port Options 1[15:14] = 0x3 (“Do Not Change”). Do not use Tx Clock Source configuration in this release. Do not use software or management tools to select a non-default Tx Clock Source. Future releases will provide full support and documented usage requirements.

1.3.1.3.4 FreeBSD Driver

• None for this release.

1.3.1.3.5 Windows Driver

• None for this release.

1.3.1.3.6 ESX Driver

• Multiple PF resets cause VMXnet3 adapter VFs in ENS mode to fail sporadically, resulting in no traffic. [18038466648]

1.3.1.4.1 General

• On Azure Stack HCI 23H2, Event Log reports Error 289 during execution of hyperv.basic.large_​send_​offload_​vf. This behavior is unexpected and is currently under investigation. [15016108905]

1.3.1.4.2 Firmware/NVM/NVM Update

• CISCO-OPLINK TRMTFEAS1CCISE1G shows link with no fiber connected. [22021216529]

• NC-SI Health Status 0x508 / 0x507 is not generated, when Global Health Status is enabled. [18042250212]

• Inconsistencies in LED behavior during system shutdown for E825-C under S5 state when using PV LEK with “NO MNG” configuration. [22021342973]

1.3.1.4.3 Linux Driver

• It is observed that forcing link speed to 1GbE for e823 with fibre optics cable causing netlink errors. [18042136652]

• It is observed that Traffic Classes for Tx queues are set to ZERO, when the DCB is configured and traffic with VLANs are sent. [16027773606]

• The dmesg log shows “link settings prevent physical link to set to OFF” when unplugging multiple cables at once. This issue seems to be limited to SFP and is under investigation. [22020900546]

• Intermittent PTP “send sync failed” can occur in Boundary Clock mode due to host-side timestamp delivery delays (observed with OS workqueue contention). [16030313929]

  Workaround: Use host configurations that minimize workqueue/graphics-driver interference during PTP runs, monitor for mgag200/EDID-related CPU hogging, and avoid affected host setups for timing-critical BC validation.

1.3.1.4.4 ESX Driver

• No traffic (ICMP, tcp, udp) after VF reset with NSX-T and DS configuration. [18041512300]

• After enabling PCI passthrough on different adapter there is no traffic. [18041539956]

1.3.1.5.1 General

• Large number of UDP threads can cause traffic to go to very low throughput. [22020878880]

1.3.1.5.2 Firmware/NVM/NVM Update

• The firmware aligns to DMTF spec clarification. The port schema defines that all the LLDPRecieve sub-properties shall not be present in the resource if the property was not received. [15019109850]

• Sysfs ‘ echo 1 > /sys/bus/pci/devices/<PCI BUS address>/remove’ command might cause the order the devices probed to change. If this is the case, the subsequent ice driver probe attempts might end up in kernel WARNING around PTP initialization, if removed PCI device is PF0. PF0 is usually a source timer owner for the devices handled by ice driver (like E825C or E830 devices) and it is a control PF for other PFs. So, non-PF0 devices require PF0 to be initialized first for a successful initialization. [16027730521]

  Workaround: If the current device probe order is disturbed, the user needs to remove all other PFs as well (PF0 as the first one) and rescan PCI bus then.

1.3.1.6.1 Linux Driver

• When certain DCB settings are enabled on E835 CQDA2 adapters, LAN and RDMA traffic may stop working, including on bonded interfaces. [21045537059]

• When an Intel E835 adapter is configured in 8×25G breakout mode, some 802.3ad bonds may fail to aggregate both 25GbE links and report 25Gb/s instead of 50Gb/s. [16031579155]

• In E835 LAG active-active configurations, RoCEv2 traffic may trigger driver hang. [15019771341]

• PTP synchronization may not converge at 200 GbE. On Intel Ethernet Controller E835-CC QSFP adapters operating in 200 GbE mode, PTP time synchronization in slave mode may fail to converge. This can prevent the system clock from stabilizing to the selected PTP master. PTP synchronization operates as expected at 25 GbE, 50 GbE, and 100 GbE. [22022048555]

• There is issue with changing msix_​eth value via devlink param interface - some settings of it on E830 cards cause bandwidth degradation. [22021080345]

• Flow Director rules are not working properly on Geneve Tunnel. Traffic is redirected to random queue. [16024525933]

• When CRC strip and VLAN strip co-exist, the user may observe a call trace in dmesg. [16025390212]

1.3.1.6.2 FreeBSD Driver

• Setting E830 and XXVDA2 OCP to only advertise 10G triggers AHS messages about failed link attempts before link comes up. [14022546016]

• ucmatose fails with RDMA_​CM_​EVENT_​UNREACHABLE (-60) on XXVDA2 OCPwhen running with -c 15000 connections. [18040525298]

1.3.1.6.3 RDMA Driver

• None for this release.

1.3.1.6.4 ESX Driver

• When loading a DDP package in ENS or ENS Interrupt mode, a warning log and intermittent ping failures may be observed. [16029517816]

  Workaround: Adding a 1-minute delay after loading the DDP package resolves the issue.

• <VMWARE> When instantiating the maximum number of VFs in NSX-T, adding a Transport Node afterwards might fail due to timeout. [18037659908]

1.3.1.6.5 Windows Driver

• TX counters are missing on Dell switches when PFC is enabled. Retrace counters are observed on hosts during testing, even with different PFC bandwidth allocations (50/50 split) [18041177557]

1.3.1.6.6 QV Tools

• Switching the port option from 8x25 to 2x1x100 using EPCT causes a BSOD approximately 2 minutes after the command is issued, resulting in a shutdown. [22022712629]

  Workaround: Reboot the system immediately after changing the port option to prevent this. Other port option transitions do not trigger the BSOD.