Fujitsu M10-1/M10-4/M10-4S Error: 'PCI access error' (12bb0000) message is reported on Fujitsu M10 systems

Asset ID:	1-72-2211342.1
Update Date:	2018-04-05
Keywords:

Solution Type Problem Resolution Sure

Solution 2211342.1 : Fujitsu M10-1/M10-4/M10-4S Error: 'PCI access error' (12bb0000) message is reported on Fujitsu M10 systems

Applies to:

Fujitsu M10-4 - Version All Versions and later
Fujitsu M10-1 - Version All Versions and later
Fujitsu M10-4S - Version All Versions and later
Oracle Solaris on SPARC (64-bit)

Symptoms

There is a case that a 'PCI access error' message is reported after Oracle Solaris boots up.
The error is due to a bug in the FRU and MSG-ID fields in the XSCF `showlogs error -V`.

In `showlogs error -V` output:
FRU: field is one of the /BB#x/CMUL, /BB#x/CMUU, /BB#x/CMUL, /BB#x/CMUU,/MBU
MSG-ID: field is one of the PCIEX-8000-YJ, PCIEX-8000-KP, PCIEX-8000-J5

Example:

XSCF> showlogs error -V
Date: xxx xx xx:xx:xx.xxx xxx xxxx
Code: 40000000-00a20400480400a204-12bb00000000000000000000
Status: Warning Occurred: xxx xx xx:xx:xx.xxx xxx xxxx
FRU: /BB#0/CMUL
Msg: PCI access error
Diagnostic Code:
    00000200 00000000 0000
    00000100 00000000 0000
    00000200 00000000 0000
    00000000 00000000 00000000 00000000
    00000000 00000000 0000
Diagnostic Messages
IO-FaultReport:
TIME            UUID
xxx xx xx:xx:xx xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
MSG-ID
PCIEX-8000-YJ
nvlist version: 0
      version = 0x0
      class = list.suspect
      uuid = xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
      code = xxxxxxxxxxxxx
      diag-time = xxxxxxxxxx xxxxxx
      de =(embedded nvlist)
      nvlist version: 0
              version = 0x0
              scheme = fmd
              authority =(embedded nvlist)
              nvlist version: 0
                      version = 0x0
                      product-id = ORCL,SPARC64-X
                      server-id = xxxxxx
              (end authority)
              mod-name = eft
              mod-version = 1.16
      (end de)

When all of the following conditions are met, this problem will occur.

Using Fujitsu M10-1/M10-4/M10-4S
CPU is SPARC64 X+ (*)
Poweron/reboot OVM control domain or OVM I/O root domain

(*) SPARC64 X+ can be identified by the XSCF showhardconf command.
       If Type field is '0x20', the CPU is SPARC64 X+.

        CPU#0 Status:Normal; Ver:4142h; Serial:xxxxxxxx;
        + Freq:3.400 GHz; Type:0x20;
        + Core:16; Strand:2;

The frequency of this problem is very rare.

Cause

This problem is caused by an XSCF firmware bug from the first XCP release.
Because the PCI Express correctable error threshold in the XSCF firmware is small, PCI access error is logged if an excess of PCI correctable errors occurs.

Solution

This bug is fixed by XCP2321. Apply XCP2321 or above to avoid this problem.

Workaround:

No workaround is available.

There is no need to replace the FRU.
If the error occurs in XCP versions effected by this bug, the solution to the problem is to clear the error.

Execute the XSCF 'showlogs error -V' command to confirm Solaris domain hostname and UUID from output. Solaris domain hostname and UUID can be confirmed by 'server-id =' field and 'uuid =' field respectively. Then ask Solaris domain administrator to execute the Solaris 'fmadm repair' command with this UUID. After 'fmadm repair' is executed, reboot the Solaris domain.

XSCF> showlogs error -V
Date: xxx xx xx:xx:xx.xxx xxx xxxx
Code: 40000000-00a20400480400a204-12bb00000000000000000000
Status: Warning Occurred: xxx xx xx:xx:xx.xxx xxx xxxx
FRU: /BB#0/CMUL
Msg: PCI access error
Diagnostic Code:
    00000200 00000000 0000
    00000100 00000000 0000
    00000200 00000000 0000
    00000000 00000000 00000000 00000000
    00000000 00000000 0000
Diagnostic Messages
IO-FaultReport:
TIME             UUID
xxx xx xx:xx:xx xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
MSG-ID
PCIEX-8000-YJ
nvlist version: 0
      version = 0x0
      class = list.suspect
      uuid = xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
      code = xxxxxxxxxxxxx
      diag-time = xxxxxxxxxx xxxxxx
      de =(embedded nvlist)
      nvlist version: 0
              version = 0x0
              scheme = fmd
              authority =(embedded nvlist)
              nvlist version: 0
                      version = 0x0
                      product-id = ORCL,SPARC64-X
                      server-id = (Solaris domain hostname)
              (end authority)
              mod-name = eft
              mod-version = 1.16
      (end de)

References

<NOTE:1533645.1> - M10-io.domain.fma - FMA running on a domain detects a PCI Express-related failure, and forwards information to XSCF
<NOTE:1924028.1> - Fujitsu M10-4/M10-4S: PCI access errors (12bb0000) on both CMUL and CMUU

Attachments

This solution has no attachment