Failing Processor Might Result in Panic Hang (24973531)

Language:

A server containing a failing SPARC M8 processor might cause the server to hang and panic. If, after the panic, the fmdump -e command displays L2 and L3 cache ereport messages for the failing processor similar to the following example, perform the actions listed in Recovery below.

# fmdump -e
. . .
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l3data-return@/SYS/CMIOU7/CM/CMP/SCC6/CORE1
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l2linefill@/SYS/CMIOU7/CM/CMP/SCC6/CORE1/L2D
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l3data@/SYS/CMIOU7/CM/CMP/L3BANK7
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l2linefill@/SYS/CMIOU7/CM/CMP/SCC5/CORE3/L2D
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l3data-return@/SYS/CMIOU7/CM/CMP/SCC6/CORE1
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l2linefill@/SYS/CMIOU7/CM/CMP/SCC6/L2I
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l3data-return@/SYS/CMIOU7/CM/CMP/SCC6/CORE1
2017-07-10/00:44:16  ereport.cpu.generic-sparc.l2linefill@/SYS/CMIOU7/CM/CMP/SCC7/CORE2/L2D
. . .

In this example, the L3 cache (l3data) and L2 cache (l2linefill) ereport messages were generated from the same processor found on an SPARC M8-8 server CMIOU installed in CMIOU slot 7.

Recovery: Replace the motherboard, processor module, or CMIOU containing the failing SPARC M8 processor.

Workaround: If you must use the SPARC T8 server or SPARC M8-8 PDomain containing the failing SPARC M8 processor, disable the SPARC M8 processor prior to powering the host back on. At the Oracle ILOM prompt, set the processor's requested_config_state property to disabled, and then power on the host.

On a SPARC T8-1 server, replace the motherboard.
On a SPARC T8-2 server, replace the processor module or disable it as follows:
```
-> set /SYS/MB/CMx/CMP requested_config_state=disabled
```
Where x is the number of the processor (0 or 1).
On a SPARC T8-4 server, replace the processor module or disable it as follows:
```
-> set /SYS/PMx/CMy/CMP requested_config_state=disabled
```
Where x is the number of the processor module (0 or 1) and y is the number of the processor (0 or 1).
On a SPARC M8-8 server, replace the CMIOU containing the failing processor or disable it as follows:
```
-> set /SYS/CMIOUx/CM/CMP requested_config_state=disabled
```
Where x is the number of the CMIOU containing the failing processor (0 to 7).