Skip to main content
  1. Home >
  2. Products >
  3. Computing Products >
  4. Servers >
  5. Fujitsu SPARC servers >
  6. Downloads >
  7. User Manuals >
  8. Fujitsu M10/SPARC M10 Systems Product Notes for XCP Version 2410 >
  9. Problems That Might Occur With Oracle Solaris 11 and Workarounds

Problems That Might Occur With Oracle Solaris 11 and Workarounds


Problems That Might Occur With Oracle Solaris 11 and Workarounds
The following table lists the problems that might occur with Oracle Solaris 11 and workarounds for them.
Table 3-49  Problems That Might Occur With Oracle Solaris 11 and Workarounds
Bug 29652141
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If you perform the live migration of a guest domain and then perform the live migration again to a system with different CPUs without rebooting its OS (*1), the migrated guest domain's clock runs wrongly (*2).
This phenomenon occurs when the guest domain to be migrated is running Oracle Solaris 11.4, or Oracle Solaris 11.3 with SRU 11.3.29.5.0 or later.
*1 The second and subsequent live migrations lead to the problem.

*2 Live migration from SPARC M10 systems to SPARC M12 systems causes the migrated guest domain's clock to run too fast. Live migration from SPARC M12 systems to SPARC M10 systems causes the migrated guest domain's clock to run too slow, which resets or panics the guest domain.
Workaround This has been modified with SRU 11.4.12.5.0. Apply SRU 11.4.12.5.0 to the guest domain to be migrated.
[How to restore]
Restart the guest domain.
Bug 28856511
29033441
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description In an environment with Oracle Solaris 11.4 or later installed, the system might hang when a 6 Gbps SAS HDD or SSD is disconnected without any OS operation performed.
The following mpt_sas driver messages may be output to the console and /var/adm/messages when these symptoms occur.
[Message output example 1]
  scsi: [ID 107833 kern.notice] /pci@9100/pci@4/pci@0/pci@0/scsi@0 (mpt_sas4):
           mptsas_ioc_task_management command completion failed
[Message output example 2]
  WARNING: /pci@8000/pci@4/pci@0/pci@0/scsi@0 (mpt_sas0):
           mptsas_check_task_mgt: Task 0x3 failed. Target=XX
[Message output example 3]
  scsi: WARNING: /pci@8000/pci@4/pci@0/pci@0/scsi@0 (mpt_sas0):
           Can only start 1 task management command at a time
Workaround This has been modified with SRU 11.4.13.4.0.
Bug 28059901
28072669
28410937
28418530
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description In an environment with Oracle Solaris 11.3 SRU 11.3.31.6.0 or later or Oracle Solaris 11.4 or later installed, the system might hang during an internal disk failure, a failure of the target 6 Gbps SAS-connected device (disk or tape), or active replacement operations.
The following mpt_sas driver messages may be output to the console and /var/adm/messages when these symptoms occur.
[Message output example 1]
  scsi: [ID 107833 kern.notice] /pci@9100/pci@4/pci@0/pci@0/scsi@0 (mpt_sas4):
            mptsas_ioc_task_management command completion failed
[Message output example 2]
  WARNING: /pci@8000/pci@4/pci@0/pci@0/scsi@0 (mpt_sas0):
            mptsas_check_task_mgt: Task 0x3 failed. Target=XX
Workaround This has been modified with SRU 11.4.2.2.0 and LSU 11.3.36.7.0.
Bug 16312511
Model SPARC M12-2S, SPARC M10-4S
Description In the environment described below, if the system board is deleted by the physical partition dynamic reconfiguration (PPAR DR) function, a system board deletion failure, stop of virtual function (VF) communication, a root domain panic, and other phenomena may occur. The environment has an SR-IOV configuration, and either a 10 Gigabit Ethernet card running the ixgbe driver for the root domain or the onboard LAN of the SPARC M12-2S is assigned there.
Workaround PPAR DR cannot be used in the above SR-IOV configuration. Use it in a virtual network (vnet) configuration.
Bug 25262938
25954620
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S
Description Execution of the nvmeadm or fwupdate command may fail with the following message displayed.
[Display example]
# nvmeadm
ERROR: Platform not supported
# fwupdate
ERROR: Platform not supported
Workaround Perform the following procedure.
1. Delete the configuration information files in /etc/ssm/hmp.

        # rm /etc/ssm/hmp/*
2. Execute nvmeadm or fwupdate.

        # nvmeadm 
                Or
        # fwupdate

   Configuration information files are automatically created in /etc/ssm/hmp.

3. Create the following configuration information files.
Create the following two files in a text editor:

    host_profile.xml
    platforms.json

- Creating host_profile.xml

# cat /tmp/host_profile.xml
<‌?xml version="1.0" encoding="utf-8"?>‌
<‌hmp_profile schemaVersion="1">‌
    <‌control>‌
        <‌!-- Time in seconds from Epoch. -->‌
        <‌timestamp>‌1438633604<‌/timestamp>‌

        <‌!-- Minimum version of HMP supported -->‌
        <‌hmp_min_ver>‌2.3.2<‌/hmp_min_ver>‌

        <‌!-- ilom, host, patch, other -->‌
        <‌origin>‌ILOM<‌/origin>‌

        <‌supported_platforms>‌SPARC M12-1<‌/supported_platforms>‌
        <‌supported_platforms>‌SPARC M12-2<‌/supported_platforms>‌
        <‌supported_platforms>‌SPARC M12-2S<‌/supported_platforms>‌
    <‌/control>‌

    <‌tool_support>‌
        <‌element>‌
 
            <‌platform_name>‌SPARC M12-1<‌/platform_name>‌
            <‌platform_name>‌SPARC M12-2<‌/platform_name>‌
            <‌platform_name>‌SPARC M12-2S<‌/platform_name>‌
            <‌utility_property>‌
                <‌name>‌ALL<‌/name>‌
                <‌supported>‌NO<‌/supported>‌
                <‌add_runtime_mode>DEFAULT<‌/add_runtime_mode>‌
            <‌/utility_property>‌
            <‌utility_property>‌
                <‌name>‌FWUPDATE<‌/name>‌
                <‌supported>‌YES<‌/supported>‌
                <‌add_runtime_mode>DEFAULT<‌/add_runtime_mode>‌
            <‌/utility_property>‌
            <‌utility_property>‌
                <‌name>‌NVMEADM<‌/name>‌
                <‌supported>‌YES<‌/supported>‌
                <‌add_runtime_mode>‌DEFAULT<‌/add_runtime_mode>‌
            <‌/utility_property>‌
        <‌/element>‌
    <‌/tool_support>‌
    <‌platform_data>‌
    {
        "version": "1.0.0.0",
        "platforms": {
            "aliases": {
                "SPARC M12-1": "undefined",
                "SPARC M12-2": "undefined",
                "SPARC M12-2S": "undefined"
            },
            "hosts": {
                "undefined": {
                    "internal_hbas": {
                        "1000:0050:1000:3060":
                        {
                            "backplanes": [ "internal_4_square" ],
                            "nac_name_format": "/SYS/SASBP/HDD%d"
                        }
                    }
                }
           }
        }
    }
    <‌/platform_data>‌
<‌/hmp_profile>‌

- Creating platforms.json

# cat /tmp/platforms.json
    {
        "version": "1.0.0.0",
        "platforms": {
            "aliases": {
                "SPARC M12-1": "undefined",
                "SPARC M12-2": "undefined",
                "SPARC M12-2S": "undefined"
            },
            "hosts": {
                "undefined": {
                    "internal_hbas": {
                        "1000:0050:1000:3060":
                        {
                            "backplanes": [ "internal_4_square" ],
                            "nac_name_format": "/SYS/SASBP/HDD%d"
                        }
                    }
                }
           }
        }
    }

4. Replace files with these configuration information files.

    # rm /etc/ssm/hmp/host_profile.xml
    # rm /etc/ssm/hmp/platforms.json
    # cp /tmp/host_profile.xml  /etc/ssm/hmp/host_profile.xml
    # cp /tmp/platforms.json /etc/ssm/hmp/platforms.json

5. Execute the fwupdate or nvmeadm command again.
No error message will appear, and the command will be executed normally.
Bug 24819902
Model SPARC M12-2S, SPARC M10-4S
Description In the control domain with SRU 11.3.8.7.0 (Oracle VM Server for SPARC 3.4) or later applied, the deleteboard command may behave inappropriately where at least one logical domain binds a core shared with another logical domain. This is because Oracle VM Server for SPARC wrongly calculates the number of CPU cores to be removed.
[Example]
- The deleteboard -m unbind=resource command succeeds but more CPUs are removed than necessary.

- The deleteboard -m unbind=resource command fails even if the CPU allocation is the same as one where it succeeds with Oracle VM Server for SPARC 3.3.

- The deleteboard -m unbind=none command fails even though there are sufficient free CPUs in the remaining system boards. 
Workaround Apply either of the following (1) or (2).
(1) Assign virtual CPUs with the whole-core constraint.

(2) Change the PPAR DR policy to "targeted".
# svcadm disable ldmd
# svccfg -s ldmd setprop ldmd/fj_ppar_dr_policy=targeted
# svcadm refresh ldmd
# svcadm enable ldmd 
Bug 20882700
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description After you execute dynamic reconfiguration of physical partitions, or after you execute the ldm add-io/remove-io command to add/delete a resource dynamically, a message indicating that the resource was added/deleted may be output to the console.
[Example 1]
SUNW-MSG-ID: FMD-8000-CV, TYPE:
Alert, VER: 1, SEVERITY: Minor
EVENT-TIME: Mon May 11 20:04:48 JST 2015
PLATFORM: ORCL,SPARC64-X, CSN:
2081232009, HOSTNAME: 4S-408-D0
SOURCE: software-diagnosis, REV: 0.1
EVENT-ID: 76d8e4f6-d621-4ede-a86e-
93abcdc908a6
DESC: FRU '/SYS//BB0/CMUU/CMP1/
MEM17A' has been removed from the
system.
...

[Example 2]
SUNW-MSG-ID: FMD-8000-A0, TYPE:
Alert, VER: 1, SEVERITY: Minor
EVENT-TIME: Thu May 14 15:50:31 JST 2015
PLATFORM: unknown, CSN: unknown, HOSTNAME: iodom0
SOURCE: software-diagnosis, REV: 0.1
EVENT-ID: 1f8d1ae8-9097-4204-b5d6-d605aac28390
DESC: FRU '/SYS/BB#1-PCI#6' has been added to the system.
...
Workaround Before executing dynamic reconfiguration of physical partitions, or before executing the ldm add-io/remove-io command to add/delete a resource dynamically, add the setting below to the /usr/lib/fm/fmd/plugins/fru-monitor.conf file of all logical domains where Oracle Solaris 11.3 is installed. Then, restart the logical domains.
setprop fmon_alert_enable "false"
Bug 20950622
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-4, SPARC M10-4S
Description Suppose the root domain is running Oracle Solaris 11.3 or later when an I/O domain created in one of the following procedures is started/restarted. Then, OpenBoot PROM aborts, and the I/O domain cannot be started/restarted.
[Creation pattern 1]
1. Add a PCI expansion unit to the root domain by using the PCI hot plug (PHP) function.

2. Add two PCI cards to the PCI expansion unit by using PHP.

3. Dynamically create a virtual function (VF) from the physical function (PF) of each PCI card.

4. If the created VF is dynamically assigned to an active I/O domain, OpenBoot PROM will abort when the I/O domain restarts.

Or
4. If the created VF is statically assigned to an inactive I/O domain, OpenBoot PROM will abort when the I/O domain starts.

[Creation pattern 2]
1. Add the PCIe bus to the root domain by using dynamic PCIe bus assignment.

2. Add two PCI cards to the same PCIe bus on the root domain by using PHP.

3. If each PCI card is dynamically assigned to an active I/O domain, OpenBoot PROM will abort when the I/O domain restarts.

Or
3. If each PCI card is statically assigned to an inactive I/O domain, OpenBoot PROM will abort when the I/O domain starts.

After step 2, even if you dynamically create a virtual function (VF) from the physical function (PF) of each PCI card and assign the created VF to an I/O domain, the same phenomenon occurs.
[Creation pattern 3]
1. With two PCI cards mounted to the same PCIe bus, use dynamic PCIe bus assignment to add the PCIe bus to the root domain.

2. If each PCI card is dynamically assigned to an active I/O domain, OpenBoot PROM will abort when the I/O domain restarts. If each PCI card is statically assigned to an inactive I/O domain, OpenBoot PROM will abort when the I/O domain starts.

After step 1, if you dynamically create a virtual function (VF) from the physical function (PF) of each PCI card and assign the created VF to an I/O domain, the same phenomenon occurs.
Workaround To prevent OpenBoot PROM from aborting, delete the VFs and the PCI cards assigned to the I/O domain by using the ldm remove-io command before starting/restarting the I/O domain. After starting the I/O domain, dynamically reassign the VFs and the PCI cards with the ldm add-io command.
Bug 21849217
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description For Oracle Solaris 11.2 SRU 11.2.14.5.0 or later and for Oracle Solaris 11.3 or later, the following error message may be output when the fwupdate command is executed, and the command may terminate abnormally.
ERROR: Could not identify host type
Workaround Delete all the files in the /etc/ssm/hmp directory, and then execute the fwupdate command.
# cd /etc/ssm/hmp
# rm -f *
Do not delete the /etc/ssm/hmp directory at this time.
After you execute the fwupdate command, new files are created in the /etc/ssm/hmp directory. Do not delete these files.
Bug -
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description When the fwupdate command is executed, the following message may be output.
Get Device ID command failed: Unknown (0x7E)
No metadata provided, so version verification can not be completed
Workaround Ignore this error message because there is no impact on command operation.
Bug 25473170
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description In Oracle Solaris 11.2 SRU 8.4 or later, suppose that the network interface of the dual 10Gbps FCoE card that is made by QLogic and runs the qlcnic driver is deactivated. When it is activated again (*1), panic may occur with the following message displayed during communication using this interface.
panic[cpux]/thread=xxxxxxxxxxx: Fatal error has occured in: PCIe fabric.(0x40)(0x41)

*1 This applies to the following cases, where the dual 10Gbps FCoE card is used.
- To change the configuration, the interface is deactivated by the ipadm delete-ip or ifconfig unplumb command, and activated by the ipadm create-ip or ifconfig plumb command.
- The FCoE interface is assigned to a logical domain. The physical partition dynamic reconfiguration (PPAR DR) function is used for continuing operation and use even after the logical domain is suspended and resumed.
- PRIMECLUSTER Global Link (NIC switching method) is used to switch or switch back a redundant NIC.
Workaround Perform the following procedure.
1. Copy the /kernel/drv/qlcnic.conf file to the /etc/driver/drv directory.

2. Add the "enable_multi_rds_sets = 0;" line to the copied file.

3. Restart the OS.
Bug -
Model SPARC M12-2S, SPARC M10-4S
Description In an environment with Oracle VM Server for SPARC 3.4 (SRU 11.3.8.7.0) or later installed on the control domain, the deleteboard(8) command may fail if all of the following conditions are met.
[Problem occurrence conditions]
1) The number of logical domains is larger than the number of existing CPU cores in the physical partition after the deleteboard(8) command is executed.
2) A logical domain is partially allocated two or more CPU cores.
3) The physical partition was dynamically reconfigured by the deleteboard(8) command with the -m unbind=resource option specified.
[Example]
XSCF> deleteboard -y -v -c disconnect -m unbind=resource 01-0
PSB#01-0 will be unconfigured from PPAR immediately. Continue?[y|n] :y
Start unconfigure preparation of PSB. [1200sec]
 :
end
PSB#01-0 could not be unconfigured from PPAR-ID 0
due to operating system or Logical Domains Manager error.
Workaround Change the PPAR DR policy to "targeted".
[Example]
primary# svcadm disable ldmd
primary# svccfg -s ldmd setprop ldmd/fj_ppar_dr_policy=targeted
primary# svcadm refresh ldmd
primary# svcadm enable ldmd
Bug -
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description A logical domain may panic if all of the following conditions are met.
1) A logical domain is running SRU 11.3.12.4.0 or later.
2) A CPU chip is degraded.
3) A logical domain is rebooted.
[Panic message example]
panic[cpuxxx]/thread=xxxxxxxxxxxx: Could not compute mpo_lgroup[] 
Workaround There is no effective workaround.