Symptom

All ThinkSystem servers listed in this document may experience intermittent hardware errors and system hangs due to an issue with the on-board Matrox MGA200 graphics chipset. This issue can be corrected by installing the required graphics device driver. Refer to the fix below for specific instructions for each OS and version.

Affected Configurations

The system may be any of the following Lenovo servers:

• ThinkSystem SN550, any model 7X16
• ThinkSystem SN850, any model 7X15
• ThinkSystem SR850, any model 7X18, 7X19
• ThinkSystem SR950, any model 7X11, 7X12
• ThinkSystem SD530, any model 7X20
• ThinkSystem SR530, any model 7X07, 7X08
• ThinkSystem SR550, any model 7X03, 7X04
• ThinkSystem SR630, any model 7X01, 7X02
• ThinkSystem SR650, any model 7X05, 7X06
• ThinkSystem ST550, any model 7X09, 7X10

The system is configured with at least one of the following:

• CentOS, any model See below for specific versions
• Oracle Enterprise Linux, any model See below for specific versions
• RHEL 6, See below for specific versions
• RHEL 7, See below for specific versions
• SLES 11, See below for specific versions
• SLES 12, See below for specific versions
• Windows Server 2012 R2, any model, any version
• Windows Server 2016, any model, any version

This tip is not option specific.

The Graphics device driver for the Matrox onboard video is affected.

The system has the symptom described above.

Note: This does not imply that the network operating system will work under all combinations of hardware and software.
Please see the compatibility page for more information:

http://www.lenovo.com/us/en/serverproven/

Solution

- SLES 11 SP4, SLES 12 SP2, RHEL 6.9, RHEL 7.3, Windows Server 2012 R2, and Windows Server 2016:

1. Go to https://support.lenovo.com/us/en/

2. Under Select a Product Family, click on Data Center then select Servers.

3. In the Select Series drop down box, select ThinkSystem.

4. In the Select Subseries drop down box, select the model of the target server.

5. In the Select Machine Type drop down box, select the machine type of the target server.

6. On the next page, select Drivers and Software from the blue banner.

7. In the Components drop down box, select Display and Video Graphics.

8. In the Operating System drop down box, select the Operating System and version.

9. This will display a list of the most recent Matrox Graphics device drivers for the target system.

10. Follow the Readme.txt file in the download package for installation instructions.

- An alternative solution for SELS 11 SP4, and SLES 12 SP2:

Install the OS using the Lenovo ThinkSystem Bootable Driver Kit for SUSE Linux Enterprise Server 11 SP4 or 12 SP2 (also known as the kISO image). Users can obtain the latest kISO image for their system from

https://drivers.suse.com/index.php?vendor=Lenovo&SLE=none&name=ThinkSystem

- CentOS 7.3 and an alternative RHEL 7.3:

Install a required update for the on-board Matrox Video Driver per Red Hat product advisory RHEA-2017:1730:

https://access.redhat.com/errata/RHEA-2017:1730

- Oracle Linux:

There is a known issue with the Matrox driver in UEK 4.1.12-94.x.x and this driver should be blacklisted by adding the following kernel parameter:

modprobe.blacklist=mgag200

Additional Information

The systems identified in the "Affected items" section running any of the specific OSes identified in the "FIX" section can experience intermittent system hangs that are the result of a synchronization issue in the on-board Matrox G200e graphics controller integrated inside the Lenovo XClarity Controller (XCC).

This issue is only exposed when a non-zero value is written to the high priority register for the G200e. When this register contains a non-zero value, a meta-stable condition is created that can randomly cause the video data queues associated with the PCIe interface to become blocked, leading to a PCIe bus completion timeout error that in turn propagates to an IERR and a system hang.

The fix is to modify the graphics driver to ensure that no value other than zero is written into the high priority register. This eliminates the possibility of the meta-stable condition from occurring and prevents the IERR from happening.