Sun Fire E25K Server
 
Sun Fire E20K Server
 
Sun Fire 12K Server
 
Sun Fire 15K Server
 

GCS>Sun Microsystems>Servers>High-End Servers
 

Applies to:

Symptoms

Cause

NOTE: Only a certified Sun Service Engineer is authorized to service the Sun Fire[TM] 12K/15K/E20K/E25K platform. Therefore if you need to obtain help in resolving a hardware problem on this platform, please open a Service Request with Sun Support Services and an engineer can assist in troubleshooting and ultimately resolving your problem.

Solution

1. Identify what the Safari Agent ID is from the message or device path in question.
2. Use the Cheat Sheets at the bottom of this document to translate the Agent ID to a physical location.

Some example messages/paths would be as follows (Safari Agent ID is highlighted):

     /SUNW,UltraSPARC-III@62,0 
     /memory-controller@0,400000 
     /pci@5c,700000 
     /address-extender-queue@7f,0 
     /SUNW,wci@15d,0:SUNW,wci-rsm 

PCI reported ECC Errors, such as the UE example below, are types of errors where it can be useful to know the location of the detecting IO Controller. The Agent ID can be determined in the event message.

Example event:

Jan  7 07:12:13  host pcisch: [ID 284024 kern.warning] WARNING: uncorrectable error detected by pci0 (safari id 00000000.0000001c) during
Jan  7 07:12:13  host         DVMA read transaction 
Jan  7 07:12:13  host  pcisch: [ID 475334 kern.info]  Transaction was a block operation. 
Jan  7 07:12:13  host  pcisch: [ID 956438 kern.info]  dvma access, Memory safari command, address 000001a0.9a550400, owned_in not asserted. 
Jan  7 07:12:13  host  pcisch: [ID 144388 kern.info]  AFSR=48000000.1c000071 AFAR=000001a0.9a550400, 
Jan  7 07:12:13  host         quad word offset 00000000.00000000, Memory Module SB13/P0/B0 J13300 J13400 J13500 J13600 id 444. 
Jan  7 07:12:13  host  pcisch: [ID 545677 kern.info]  mtag 0, mtag ecc syndrome 0 
Jan  7 07:12:13  host  pcisch: [ID 308334 kern.info]  secondary error from DVMA read transaction 
Jan  7 07:12:13  host  unix: [ID 836849 kern.notice] 
Jan  7 07:12:13  host  ^Mpanic[cpu417]/thread=3002663b9c0: 
Jan  7 07:12:13  host  unix: [ID 261965 kern.notice] Fatal PCI UE Error 

1) The Agent ID of the implicated IOC is detailed in the message:

Jan  7 07:12:13  host         quad word offset 00000000.00000000, Memory Module SB13/P0/B0 J13300 J13400 J13500 J13600 id 444. 

When 444 is converted to hexidecimal, it is 1BC, which is the Agent ID.

NOTES:

• Make sure to know which type of System Board (SB) you are decoding. Output from the System Management Services (SMS) command showboards, as well as the domain post logs, and boot messages will show you which type of SB is installed on platform:

• USIII shows up as board type CPU

• USIV shows up as board type V3CPU

• MaxCPU boards are only supported in Sun Fire[TM] 12K/15K platforms and are USIII boards.

• The memory-controller is the CPU module on a Sun Fire[TM] 12K/15K/E20K/E25K System Board.

• A memory-controller device path will have a different Memory Offset depending on the Dimm Bank that is implicated:

• 400000 for Dimm Bank 0.

• 600000 for Dimm Bank 1.

• In the tables below, the number outside of the parenthesis is the Safari Agent ID, also called the Hexidecimal ID. In the parenthesis is the Decimal ID.

EXAMPLES:

1) Decoding an UltraSPARC III CPU:

                   /SUNW, Ultrasparc III@1e9,0 
                                          /\ 
                                       Agent ID.

Using the USIII Table below, Agent ID 1e9 is MaxCPU 1, located on Expander 15.

2) Decoding a memory bank:

                /SUNW, memory-controller@2,600000 
                                        /\    /\ 
                                   Agent ID  Memory Offset 

Using the USIII Table below (Confirmed the board type was USIII first), Agent ID 2 is CPU2 on SB0.

The Memory Offset of 600000 means the implicated dimm bank is bank 1.

3) Decoding an UltraSPARC IV CPU:

                    /SUNW, UltraSPARC-IV@1c4,0
                                          /\
                                       Agent ID

Using the USIV Table below, Agent ID 1c4 is CORE 0.1 on Expander 14.

This solution has no attachment