Sun Fire V480 and V880 With 900 MHz CPUs May Panic or "Red State" Due to Incorrect L2 SRAM Parameter Settings in Firmware |
|
| Category : | Availability |
| Release Phase : | Resolved |
| Product : | Sun Fire V880 Server Sun Fire V480 Server
|
| Bug Id : | 4772534, 4841493
|
| Date of Resolved Release : | 10-JUL-2003
|
Impact
Sun has identified an issue with L2 SRAM parameter settings on Sun Fire V480 and Sun Fire V880 systems with 900 MHz CPUs. This issue may cause L2 SRAM errors to be produced, which may cause a system to experience a panic and/or "Red State" exception error which may loop indefinitely. ("Red State" exception error messages are generated by the OBP when a CPU is hung and normal execution cannot be continued.)
Contributing Factors
This issue can occur in the following releases:
SPARC Platform
-
Sun Fire V480 with firmware 4.7.5 or earlier (delivered in patch 113034-06 or earlier)
-
Sun Fire V880 with firmware 4.7.5 or earlier (delivered in patch 112186-08 or earlier)
The current OBP firmware revision level can be obtained by running the "prtdiag -v" command.
Symptoms
This issue may cause the prtfru(1M) command to fail and depending on the version of Solaris, one of the following error messages may be experienced:
Failed to get frutree root node, libfru: IO error
This system does not provide FRU ID data
Also when this issue is present, error messages with one of the following character strings may be encountered:
UCC, UCU, EDC, EDU, WDC, WDU, CPC, CPU(not to be confused with references to a processor)
Should the "Red State" issue be encountered, messages similar to the following may be displayed on the console:
ERROR: CPU0 RED State Exception
System State (CPU0 reporting)
CPU0 Config/Control/Status registers:
CPUVersion: 003e.0014.5400.0507
SafConfig: 0caa.01bc.0000.8002
SafBaseAdr: 0000.0400.0000.0000
DCacheCtl: 0000.0000.0000.0000
ECacheCtl: 0000.0000.0009.4400
ECErrEnable: 0000.0000.0000.0000
AFAR: 0000.0040.ffc0.0d10
AFSR: 0030.0600.0000.00a0 Multiple PRIV UCC UCU DMMU SFAR: 0000.07ff.fff8.0100
DMMU SFSR: 0000.0000.0004.802c TM CT1 PR W
IMMU SFSR: 0000.0000.0080.8008 TM PR
CPU0 Trap registers: Trap Level = 5
TL=1 TT: 0000.0000.0000.0068
TSTATE: 0000.0008.7400.1600 XCC:(clear) ICC:N MM=TSO PEF PRIV
IE
TPC: 0000.0000.f004.982c
TnPC: 0000.0000.f004.9830
TL=2 TT: 0000.0000.0000.0068
TSTATE: 0000.0099.4604.1400 XCC:NC ICC:NC MM=TSO PEF PRIV
TPC: 0000.07ff.f000.d12c
TnPC: 0000.07ff.f000.d130
TL=3 TT: 0000.0000.0000.0010
TSTATE: 0000.0099.4604.1400 XCC:NC ICC:NC MM=TSO PEF PRIV
TPC: 0000.0000.0000.0004
TnPC: 0000.0000.3d09.3510
TL=4 TT: 0000.0000.0000.0010
TSTATE: 0000.0099.4600.1500 XCC:NC ICC:NC MM=TSO PEF PRIV AG
TPC: 0000.0000.f000.4200
TnPC: 0000.0000.f000.4204
*TL=5 TT: 0000.0000.0000.0010
TSTATE: 0000.0099.4600.1500 XCC:NC ICC:NC MM=TSO PEF PRIV AG
TPC: 0000.0000.f000.4200
TnPC: 0000.0000.f000.4204
The system may loop indefinitely trying to recover as shown above.
Workaround
To recover from a "Red State" loop, turn the keyswitch to diagnostic mode which should break the loop and fail out a module. The system should be power cycled next and POST rerun to verify if there is a hard failure or not. If the item does not fail on the rerun of POST, you should contact Sun Services support to confirm whether you have experienced this issue.
To verify the failure signature, the ttya console output must be logged and sent to Sun Services support.
Resolution
This issue is addressed in the following releases:
SPARC Platform
-
Sun Fire V480 with OBP firmware revision 4.10.7 (delivered in patch 113034-07) or later firmware revisions
-
Sun Fire V880 with OBP firmware revision 4.10.7 (delivered in patch 112186-09) or later firmware revisions
When upgrading the firmware, it is also a requirement to upgrade the following patches:
-
Solaris 8 with PICLD patch 110460-26 or later
-
Solaris 8 with kernel patch 108528-16 or later
-
Solaris 9 with PICLD patch 114375-01 or later
-
Solaris 9 with kernel patch 112233-02 or later
Modification History
AttachmentsThis solution has no attachment