Home EDITOR’S PICKS Intel® Ethernet Adapter Known Problems and Solutions

Intel® Ethernet Adapter Known Problems and Solutions

by Cengiz Kuskaya

Mentioned below some of the most known Intel® Ethernet Adapter Problems and Solutions provided by Intel.

1. Unable to shutdown virtual machine.
Multiple VF failover events may leave a VM in an unstable state. You may not be able to shutdown the VM. Rebooting the host will resolve the issue.

2. Traffic does not transmit through VXLAN tunnel.
On a system running Microsoft* Windows Server* 2016, traffic may fail to transmit through a VXLAN tunnel. Enabling Transmit Checksum Offloads for the appropriate traffic type will resolve the issue. For example, set “TCP Checksum Offload (IPv4)” to “TX Enabled” or “RX & TX Enabled”

3. hv_vmbus probe error on a Linux guest in a Windows Server system.
On a system running Microsoft Windows Server 2019 or Windows Server 2016 on the host and Linux in the VF, you may see an “hv_vmbus: probe failed for device X” error in dmesg after you change a vSwitch from VMQ to SRIOV. This is due to a known timing issue in the operating system. There is no functionality loss, and the VF will successfully start after a few failed probes.

4. Incomplete branding string displayed in the event log.
Some branding strings are too long to be displayed fully in the event log. In these cases, the branding string will be truncated and the port’s PCI Bus/Device/Function are appended to the string. For example: Intel(R) Ethernet Converged Network Ad… [129,0,1].

5. PcieLinkSpeed is Unknown.
When you install an Intel® Ethernet Controller 800 series device in a PCI Gen 4 slot, the operating system may report PcieLinkSpeed as Unknown. This does not affect the operation of the device.

6. Windows guest OSs on a Linux host may not pass traffic across VLANs.
The VF is not aware of the VLAN configuration if you use LBFO to configure VLANs in a Windows guest. VLANs configured using LBFO on a VF driver may result in failure to pass traffic. You must use Windows Hyper-V on the host to configure VLANs on a Windows guest.

7. Intermittent Link Loss and Degraded Performance at High Stress Can Occur on Windows Server 2012 Systems.
In a Windows Server 2012-based system with multi-core processors, possible intermittent link loss and degraded performance at high stress may occur due to incorrect RSS processor assignments. More information and a Microsoft hotfix are available at: http://support.microsoft.com/kb/2846837.

8. Virtual machine loses link on a Microsoft Windows Server 2012 R2 system.
On a Microsoft Windows Server 2012 R2 system with VMQ enabled, if you change the BaseRssProcessor setting, then install Microsoft Hyper-V and create one or more virtual machines, the virtual machines may lose link. Installing the April 2014 update rollup for Windows RT 8.1, Windows 8.1, and Windows Server 2012 R2 (2919355) and hotfix 3031598 will resolve the issue. See http://support2.microsoft.com/kb/2919355 and http://support2.microsoft.com/kb/3031598 for details.

9. DCB QoS and Priority Flow Control do not act as expected.
If you use Microsoft’s Data Center Bridging (DCB) implementation configure Quality of Service (QoS) and Priority Flow Control (PFC), the actual traffic flow segregation per traffic class may not match your configuration and PFC may not pause traffic as expected. If you mapped more than one priority to a Traffic Class, enabling only one of the priorities and disabling the others will work around the issue. Installing Intel’s DCB implementation will also resolve this issue. This issue affects Microsoft Windows Server 2012 and Server 2012 R2.

10. Link loss after changing the Jumbo Frames setting.
Inside a guest partition on a Microsoft Windows Server 2012 R2 Hyper-V virtual machine, if you change the jumbo frame Advanced setting on an Intel® X540 based Ethernet Device or associated Hyper-V NetAdapter, you may lose link. Changing any other Advanced Setting will resolve the issue.

11. Virtual Machine Queues are not allocated until reboot.
On a Microsoft Windows Server 2012 R2 system with Intel® Ethernet Gigabit Server adapters installed, if you install Hyper-V and create a VM switch, Virtual Machine Queues (VMQ) are not allocated until you reboot the system. Virtual machines can send and receive traffic on the default queue, but no VMQs will be used until after a system reboot.

12. Application Error Event IDs 789 and 790 in the Event Log.
If Data Center Bridging (DCB) is enabled, and the enabled port loses link, the following three events may be logged in the event log:

Event ID 789: Enhanced Transmission Selection feature on a device has changed to non-operational l Event ID 790: Priority Flow Control feature on a device has changed to non-operational

This is the expected behavior when a DCB enabled port loses link. DCB will begin working again as soon as link is reestablished. A port will lose link if the cable is disconnected, the driver or software package is updated, if the link partner goes down, or for other reasons.

13. “Malicious script detected” Warning from Norton AntiVirus During PROSet Uninstall.
The Intel PROSet uninstall process uses a Visual Basic script as part of the process. Norton AntiVirus and other virus scanning software may mistakenly flag this as a malicious or dangerous script. Letting the script run allows the uninstall process to complete normally.

14. Unexpected Connectivity Loss.
If you uncheck the “Allow the computer to turn off this device to save power” box on the Power Management tab and then put the system to sleep, you may lose connectivity when you exit sleep. You must disable and enable the NIC to resolve the issue. Installing Intel® PROSet for Windows Device Manager will also resolve the issue.

15. VLAN Creation Fails on a Team that Includes a Non-Intel Phantom Adapter.
If you are unable to create a VLAN on a team that includes a non-Intel phantom adapter, use Device Manager to remove the team, then recreate the team without the phantom adapter, and add the team to the VLAN.

16. Receive Side Scaling value is blank.
Changing the Receive Side Scaling setting of an adapter in a team may cause the value for that setting to appear blank when you next check it. It may also appear blank for the other adapters in the team. The adapter may be unbound from the team in this situation. Disabling and enabling the team will resolve the issue.

17. A VLAN Created on an Intel Adapter Must be Removed Before a Multi-Vendor Team Can be Created.
In order to create the team, the VLAN must first be removed.

18. RSS Load Balancing Profile Advanced Setting.
Setting the “RSS load balancing profile” Advanced Setting to “ClosestProcessor” may significantly reduce CPU utilization. However, in some system configurations (such as a system with more Ethernet ports than processor cores), the “ClosestProcessor” setting may cause transmit and receive failures. Changing the setting to
“NUMAScalingStatic” will resolve the issue.

19. Opening Windows Device Manager property sheet takes longer than expected.
The Windows Device Manager property sheet may take 60 seconds or longer to open. The driver must discover all Intel Ethernet devices and initialize them before it can open the property sheet. This data is cached, so subsequent openings of the property sheet are generally quicker.