Description

Processes on SPARC T5-x and SPARC M5-32 systems with power management policy set to either 'elastic' or 'performance' may exhibit erroneous behavior due to non-monotonically increasing time reports from gethrtime(3C) and related functions. This unpredictable behavior may also lead to system panics.

Occurrence

This issue can occur on the following systems:

• SPARC T5-1B Server with System Firmware 9.0.0.i (patch 16880751) or earlier
• Netra SPARC T5-1B Server with System Firmware 9.0.0.i (patch 16880765) or earlier
• SPARC T5-2 Server with System Firmware 9.0.0.i (patch 16880749) or earlier
• SPARC T5-4 Server with System Firmware 9.0.0.i (patch 16880761) or earlier
• SPARC T5-8 Server with System Firmware 9.0.0.i (patch 16880761) or earlier
• SPARC M5-32 Server with System Firmware 9.0.1.g (patch 16880722) or earlier

Note: No other systems or platforms are affected by this issue.

To determine the firmware version installed on the system, use the following methods:

1. Log into the Service Processor and run:

    -> show /HOST sysfw_version

        /HOST  
         Properties:
            sysfw_version = Sun System Firmware 9.0.0.i 2013/05/16 17:45

2.  From Solaris:

    # prtdiag -v | grep Firmware
    Sun System Firmware 9.0.0.i 2013/05/16 17:45

Symptoms

If the described issue occurs, the inconsistent reporting of load average may be seen by periodically running the uptime(1) command from Solaris. The example below shows instances where the 15-minute load average may show an unreasonably large value:

    7:07am  up 5 day(s), 11:44,  8 users,  load average: 3.78, 2.79, 8388608.00
    7:08am  up 5 day(s), 11:44,  8 users,  load average: 3.55, 2.80, 8388608.00
    7:08am  up 5 day(s), 11:44,  8 users,  load average: 3.54, 2.84, 8388608.00
    7:08am  up 5 day(s), 11:45,  8 users,  load average: 3.61, 2.91, 8388608.00
    7:09am  up 5 day(s), 11:45,  8 users,  load average: 3.64, 2.96, 8388608.00
    7:09am  up 5 day(s), 11:45,  8 users,  load average: 1388161.20, 286921.04, 8388608.00
    7:09am  up 5 day(s), 11:46,  8 users,  load average: 1716760.39, 450624.29, 8388608.00
    7:10am  up 5 day(s), 11:46,  8 users,  load average: 1231341.42, 421832.24, 8388608.00
    7:10am  up 5 day(s), 11:46,  8 users,  load average: 883176.39, 394879.84, 8388608.00

Processes may exhibit abnormal or unpredictable behavior due to non-monotonically increasing time from gethrtime(3C) and related functions.

If the described issue occurs resulting in a system panic, a system panic message may be seen on the host console.  Due to the non-deterministic nature of the problem, the panic does not generate reliably reproducible panic output and therefore cannot be documented here.

Workaround

To avoid this issue, turn off the server's power management feature. This can be achieved by doing both of the following:

1. From ILOM Shell:

For SPARC T5-X Server:

     -> set /SP/powermgmt policy=disabled

For SPARC M5-32 Server:

Disable power management on each physical domain:

    -> set /Servers/PDomains/PDomain_<#>/SP/powermgmt policy=disabled

where <#> is the physical domain ID; for example:

    -> set /Servers/PDomains/PDomain_0/SP/powermgmt policy=disabled

2. From Solaris 11:

    # poweradm set administrative-authority=platform

This issue is addressed in the following firmware releases:

System Firmware 9.0.2.g or later, as delivered in the following patches:

• SPARC T5-1B Server with patch 17019075 or later
•  Netra SPARC T5-1B Server with patch 17019067 or later
• SPARC T5-2 Server with  patch 17019079 or later
• SPARC T5-4 Server with  patch 17019069 or later
• SPARC T5-8 Server with patch 17019069 or later

System Firmware 9.0.2.h or later, as delivered in the following patches:

• SPARC M5-32 Server with patch 17308609 or later

Patches

<SUNPATCH:17019075>
<SUNPATCH:17019067>
<SUNPATCH:17019079>
<SUNPATCH:17019069>
<SUNPATCH:17308609>

History

19-Aug-2013: Document released - State: Resolved.
21-Oct-2013: Maintenance update; format corrections, no change in content

Internal Section:

Comments:
The following bugs are identified as having the same root cause:
17213398, 17185976, 17072620

Questions regarding this document should be addressed to
sunalertpublication_us_grp@oracle.com and copy the
responsible engineer listed below.

Internal Contributor/Submitter: marcel.widjaja@oracle.com, eugene.loh@oracle.com, julia.harper@oracle.com, matt.noel@oracle.com, bill.holler@oracle.com
Internal Eng Responsible Engineer: mark.kanda@oracle.com, janelle.tavares@oracle.com
Internal Services Knowledge Engineer: jeff.folla@oracle.com
Internal Eng Business Unit Group: Solaris Revenue Product Engineering
Internal Escalation ID: 3-7458896351
Internal Pending Patches:
Internal Resolution Patches: 17019075, 17019067, 17019079, 17019069, 17308609

References

This solution has no attachment