Intel® Virtual RAID on CPU (Intel® VROC) Limitations in Windows* Environments

When a RAID volume is created with a drive that has an operating system on it (the system drive) and then migration is in progress, if a separate RAID volume is created shortly after, this new non-system RAID volume cannot be deleted within the Intel® VROC Graphical User Interface (GUI) or Command-Line Interface (CLI) utility. The user who attempts to do so will be greeted with an error pop-up in the Intel® VROC GUI utility or with the following error message within the Intel® VROC CLI utility: Request is formatted correctly but failed to execute, within the CLI tool.

The Intel® VROC product installation application does not include Microsoft* .NET Framework. Visit Microsoft to download the latest version. Intel® VROC installation may not complete successfully without this framework. The Intel® VROC Graphical User Interface (GUI) application requires the presence of the latest version of Microsoft* .NET Framework.

Intel® VROC (VMD NVMe* RAID) has the following limitation when hot inserting an NVMe* drive:

• When hot inserting multiple drives, allow enough wait time between each drive for the platform to process each event.
• When inserting an entire RAID volume, Intel recommends that the system be shut off, the disks inserted and then the system powered back on. If an attempt is made to hot insert all of the drives of the RAID volume, the RAID volume may enter into a fail state because of the above-mentioned delay. If the RAID volume remains in a failed state after all of the drives are inserted, you can attempt to recover the RAID volume by using the Intel® VROC Graphical User Interface (GUI) to reset the volume to normal. This may allow the volume to either continue to operate, or force a volume rebuild. There is a risk that the RAID volume could become inoperable.

Due to Microsoft Windows* time restrictions for resuming from S3 and S4 power states, and Intel® Volume Management Device (Intel® VMD) identification requirements, hot-plug (insertion and/or removal) of Intel® VMD-enabled NVMe* devices is not supported during S3 and S4 power states.

If multiple NVMe* SSDs in a system need to be replaced via the hot-plug process, the user must give enough time between each process activity. Release one drive at a time. There also must be from 45 to 90 seconds (or until the device re-appears/disappears in the Microsoft* Device Manager) between hot-plug events. One hot-plug event would be the hot removal of the drive. The next hot-plug event would be the hot insertion of the new drive.

Simultaneous hot-plug events of multiple NVMe* drives are not supported. The user must wait until a device is reflected as removed/inserted in the Microsoft* Device Manager for spacing surprise hot-plug of Intel® VMD enabled PCIe NVMe* drives in Microsoft Windows*.

Due to these limitations, Intel strongly discourages performing hot-plug events during an S3/S4 power state change.

On a RAID 5 volume, disk cache is being turned off when a volume is degraded. Due to this, the rebuilding times have increased expectedly until the rebuild is completed, and disk cache is enabled again.

This extends to drives being added to a RAID 5 volume as well.

The Intel® VROC CLI utility does not support the RAID volume name beginning with blank space.

Once an Intel® VROC upgrade key has been inserted into the system, the trial version is concluded. Removing the upgrade key does not re-enable the trial version. As a result, any existing RAID volumes present while the upgrade key was installed will not be seen and could be in an unknown state.

When creating a RAID volume using the trial version, do not mix NVMe* drive vendors. Mixing vendors may result in unexpected behaviors.

Additional considerations during the Intel® VROC trial period:

• While it is possible to use bootable RAID volumes, it is highly recommended to use only data RAID volumes during the trial period. Having the operating system in a bootable RAID volume may cause users to be locked out completely from their system if the trial period runs out.
• During the trial period, it is highly recommended to use data RAID volumes on same make/model of NVMe* devices. This is the suggested configuration leading to fewer potential issues, although it is possible to create volumes with different make/model drives.

The Intel® VROC UEFI RAID drivers comply with the UEFI Specifications for PCI Driver Model for PCI Device Drivers (Section 13.3.3) and may return the status code access denied from the UninstallProtocolInterface routine from boot services (spec. 6.3). This is an expected behavior.

The use of the included Intel® VROC F6 drivers are required to install an operating system onto an Intel® VROC managed device. There is no Microsoft* inbox driver that supports Intel® VROC.

The supported Microsoft* operating systems for Intel® VROC include inbox drivers that support the Intel® C620, C422 and C741 series chipset Platform Controller Hub (PCH) when configured for RAID mode. It is strongly recommended that the Intel® VROC (SATA RAID) F6 drivers included in the Intel® VROC software package are used instead of the available Microsoft* inbox driver. The provided inbox driver is intended only for those customers who may not have the Intel® VROC (SATA RAID) F6 drivers readily available and only for installing to a single drive (not to a RAID volume). Once the operating system is installed, it is required for proper Intel® VROC support that the Intel® VROC installer package is installed immediately. At that point, it will be safe to migrate the SATA system disk into a RAID volume (using the Intel® VROC Graphical User Interface (GUI) utility).

This limitation is in reference to platforms having a Platform Controller Hub (PCH) that supports more than 6 SATA ports. The Intel® C620 and C422 series chipset SATA controller supports 8 SATA ports. As referenced above, the Windows* operating systems that contain the inbox drivers for the Intel® C620 and C422 series chipset PCH when configured for RAID mode, only support 6 ports. Drives on ports 7 and/or 8 are not enumerated. For this reason, Intel recommends not using these two ports as part of the Windows* operating system boot installation (as a pass-thru drive or as part of a RAID volume). However, if you do need to use these ports as part of your Windows* boot volume, the steps below can be used as a workaround. You will need a USB drive with the Intel® VROC Command-Line Interface (CLI) utility.

1. After you have created the desired RAID volume that includes ports 7 and/or 8 (which you intend to use as your Windows* boot volume) in the Pre-OS environment, begin the Windows* installation process. Make note of the RAID volume name.
2. Navigate to the Windows* disk selection window. At this point, select the option Load Driver and install the Intel® VROC F6 driver (included in the Intel® VROC software package).
3. Attempt to continue installing the Windows* operating system onto the RAID volume. If the installation process does not continue, the limitation described above has been encountered.
4. Press the key F10 to invoke a command prompt window.
5. If you have not already done so, insert the USB drive into the system. Navigate to your USB drive with the Intel® VROC CLI utility.
6. Run the following command: IntelVROCCLI.exe --manage --normal-volume <volumeName>. This will reset the volume to a normal state.
7. Close the command prompt window.
8. In the Windows* disk selection window, reload the Intel® VROC F6 driver. Once completed, Windows* should allow the installation on the RAID volume.

With Microsoft Windows* 10 and Windows* 11 operating systems, Fast Startup is enabled by default. Disable Fast Startup prior to removing/upgrading the Intel® VROC hardware key, or alternatively, perform a complete reboot when removing/inserting an Intel® VROC hardware key when Fast Startup is enabled.

In Windows* and UEFI, the port number shown in the Intel® VROC interfaces depends on the disk enumeration order by the Intel® Volume Management Device (Intel® VMD)-enabled NVMe* driver, which can be different on each platform. The port numbers shown do not reflect the physical PCIe slot. After each hot-plug event, there is an enumeration process that is not fixed.

Intel® Rapid Storage Technology enterprise (Intel® RSTe) 5.5.0.2013 introduces support for Windows* 10 RS5 and Windows* Server 2019.

There is a known issue trying to install Windows*10 RS5 or Windows* Server 2019. Installing Windows* 10 RS5 or Windows* Server 2019 onto an Intel® Volume Management Device (Intel® VMD)-managed device is limited to a single CPU. For more information, refer to the known issues in Intel® Virtual RAID on CPU (Intel® VROC) Release Notes for Windows*.

This limitation only applies to Intel® RSTe 5.5 and earlier releases. Intel® VROC 6.x and newer are not impacted.

Due to restrictions on NVMe* implementations that reside behind PCI compliant bridges, address space is limited to 32-bit assigned base address registers (for non-prefetchable memory). This means that there is only 4GB of addressable (non-prefetchable) memory space available. This memory space must be shared between the entire platform. As a result, there is a limitation on the number of Intel® VMD controllers that can be enabled (due to these resource constraints). The following two configurations can be used without encountering memory restriction errors:

1. Three Intel® VMD domains can be enabled on CPU0 up to CPU6. Intel® VMD domains on CPU7 need to remain disabled. This results in a maximum of 21 Intel® VMD domains that can be enabled in this configuration.
2. No more than two Intel® VMD domains can be enabled on each CPU in an 8-socket platform. This results in a total of 16 Intel® VMD domains.

This limitation does not apply to 1, 2 or 4-socket platforms.

When designing a Hot Swap Backplane (HSBP) into a new platform, make sure that the backplane design supports the platform’s external design document specification for HSBP support.

The sSATA controller on the Intel® Cooper City and Wilson City Customer Reference Boards (CRBs) have very limited support for using and/or testing backplane management. Make sure to review all Intel® Cooper City or Wilson City CRB design documentation to understand how the sSATA controller is laid out.

When creating a RAID volume, there will be a volume size difference seen when comparing a RAID volume created using the Intel® VROC Pre-OS HII environment and volumes created from the operating system. This has to do with the way the size information is displayed and computed between the two environments.

When CTO occurs, Intel® VMD must be chosen to handle these conditions when the root port of the Intel® VMD controller is enabled. This is accomplished by disabling IOMCA on the x16 Intel® VMD enabled lanes.

The NVMe* device name in the Microsoft* Device Manager is created by Microsoft*. Within the Microsoft* Device Manager, the model number of NVMe* devices may not be completely displayed once Intel® Volume Management Device (Intel® VMD) is enabled. A workaround is possible with a new parameter in UnitControl, ScsiUnitRichDescription to show all model numbers in the device name. The workaround can only solve the problem when the drives have the same model number but different sizes.

When using AtaPassThru protocol to get information on attached devices from the Intel® VROC Pre-OS, the value of GetNextTargetLun may not be successful. A workaround is to set the value of the Timeout field inside Packet to a larger value (such as 3 seconds).

Warning Message Displayed When GUI and Driver Version Mismatch

When this warning message condition is detected from one major version or two minor versions' difference, the Intel® VROC GUI utility will not allow you to manage the storage controller, but can provide a set of functionalities corresponding to the lowest driver version. The Intel® VROC GUI utility still can start even if there is no driver installed.

RAID volume creation could not be done after enabling Pre-Boot DMA Protection and DMA Control Opt-In Flag. This feature enablements are targeted to be corrected by Eagle Stream Refresh (EGS-R) timeframe.

Beginning with Microsoft Windows 11* SV2 release timeframe, Microsoft* has implemented a correctness fix to address a potential race condition that could result in a system failure (bugcheck). The issue corrected was in the device PnP removal process. This correctness fix has been backported to Microsoft Windows 11* SV from 2021.10C, Windows Server* 2022 from 2021.08C and Windows® 10 22H2. As a result of this Microsoft* correction, one of the Windows* inbox RAID drivers, the Intel® Rapid Storage Technology (Intel® RST) driver, now fails the device PnP removal process. This process failure blocks the Intel® RST driver from being properly removed. As a result, a replacement driver cannot be installed. Long-term solutions to resolve this issue have been brought up to Microsoft to enact corrective measures.

A potential silent data loss/corruption condition exists in all Intel® Rapid Storage Technology enterprise (Intel® RSTe) Windows* drivers, beginning with version 5.4.0.1465, as well as all Intel® Virtual RAID on CPU (Intel® VROC) Windows* driver versions (until Intel® VROC 7.x). This potential silent data loss/corruption condition occurs on Intel® VROC managed RAID 5 volumes, consisting of at least one drive that supports TRIM/UNMAP commands, by methods other than the use of zeros.

A potential silent data loss/corruption condition (SDC) exists in all Intel® Virtual RAID on CPU (Intel® VROC) Windows* driver versions beginning with the release package version 7.0.0.2291. The issue has been root caused to a very small timing window where a particular sequence of events occurs. A customer may be exposed to this SDC condition when all the following conditions have been met:

• The platform is configured with a RAID 5 volume, and it is in a rebuild state.
• The platform is under very heavy I/O.
• The I/O operation generates a WRITE operation to a RAID volume stripe that is not being rebuilt yet.
• The I/O operation generates a READ operation to the RAID volume, in which the READ address range overlaps with two stripes: the stripe that is being rebuilt, and the next stripe that has yet to be rebuilt.

If these conditions have been encountered in an operational environment, the platform may encounter the SDC condition. Intel strongly recommends that customers, who may be exposed to this silent data loss condition, immediately update to the Intel® VROC 8.2 release version.