Sun Storage 6920 System
 

GCS>Sun Microsystems>Sun Alert>Criteria Category>Data Loss
 
GCS>Sun Microsystems>Sun Alert>Release Phase>Resolved
 

On a Sun StorEdge 6920 array, firmware on the Storage Resource Card (SRC) contains code that may incorrectly disable interrupt handlers in the case of a specific sequence of multi-bit memory errors. This could allow incorrect data to be written to permanent storage.

This issue can occur on the following platform:

• Sun StorEdge 6920 Arrays with System Processor (SP) version 2.0.4 and earlier

The system processor version can be determined from "/etc/release" with the following command (run on the SP):

    [sp0]# cat /etc/release
    Solaris 9 s9_58shwpl3 SPARC
    ...
              Sun StorEdge(tm) 6920
                 Version: 2.0.2
            Build Date: July  2, 2004
    ...

If StorADE generates the following error message, this is an indication that the SRC card may not be working properly and/or writes to memory may fail:

    LOG_ALERT (POST: 0-0)  SRC2 slot:3, CPU:1, Sp Local Memory
    Detected Multi-Bit ECC Error: Disabling Interrupt BEWARE

There is no workaround. Please see the Resolution section.

This issue is addressed on the following platform:

• Sun StorEdge 6920 Arrays with System Processor (SP) version 2.0.5 and later

Notes:

1. An immediate upgrade for Sun StorEdge 6920 systems to SP version 2.0.5 should be scheduled in order to avoid this issue.
2. StorADE is used to upgrade all 6920 components to version 2.0.5. For more information on all 6920 procedures, please see the 2.0.5 Release Notes at http://www.sun.com/products-n-solutions/hardware/docs/pdf/817-5229-14.pdf.

Normally, when handling a multi-bit ECC error, the storage processor experiencing the problem is reset.

In rare circumstances, if a particular multi-bit ECC error occurs, the SE6920 disables the interrupt handler. Subsequent reads or writes to defective memory may result in incorrect data being written to permanent storage because the interrupt handler will no longer reset the DSP when multi-bit errors are detected.

This solution has no attachment