Skip to main content

Problems Resolved in Oracle Solaris 11.3


Problems Resolved in Oracle Solaris 11.3
The following table lists the problems resolved in Oracle Solaris 11.3. You might encounter them in supported releases earlier than Oracle Solaris 11.3.
Table 3-52  Problems Resolved in Oracle Solaris 11.3
Bug 15813959
15813960
(7196117)
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description A PCI expansion unit is added using Oracle Solaris hotplug(1M) in a SPARC M12/M10 system. However, devices on the PCI expansion unit are not recognized.
Workaround Before you add a PCI expansion unit by hotplug(1M), add the following line in the /etc/system file in advance and restart Oracle Solaris.
set pcicfg:pcicfg_slot_busnums = 4

Note that the system does not recognize a device of a PCI expansion unit if you add the PCI expansion unit by PHP to a root complex that has been added by either of the following: the dynamic reconfiguration of the physical partition, or the Dynamic PCIe bus assignment.
If this problem occurs, restart the logical domain to which the PCI expansion unit is assigned to make the system recognize the device of the PCI expansion unit.
Bug 17430911
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S
Description When you change the power-saving operation of the physical partition from "elastic" to "disabled," the CPU frequency assigned to the logical domain may not increase.
Workaround This has been modified with SRU 11.2.8.4.0 (Oracle VM Server for SPARC 3.2).
Install SRU 11.2.8.4.0 or later on the control domain.
[How to restore]
Execute the Oracle Solaris svcadm command in the control domain to restart the ldmd services.
primary# svcadm restart ldmd
Bug 17561541
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description In a SPARC M10 environment with XCP 2230 or later applied, or in a SPARC M12 environment with XCP 3021 or later applied, suppose that the ldm add-io command is executed after the ldm remove-io command was executed during delayed reconfiguration. Then, the ldmd daemon may cause a core dump and restart.
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
During delayed reconfiguration, execute the ldm remove-io command after executing the ldm add-io command.
Bug 18502702
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If the SunVTS 7.0 ps17. 1 test is started on a SPARC M10 system with SPARC64 X+ processors, it may end with an error.
Workaround This has been modified with SRU 11.2.1.5.0 and patch 151265-03 for Oracle Solaris 10.
Bug 18595023
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If the ldm list-io command is executed after PCI cards, which support the SR-IOV function, are mounted on PCI expansion unit SLOT4 or higher, the pseudonym of the physical function of the PCI cards mounted in SLOT4 or higher is mistakenly shown as SLOT2. Moreover, the virtual functions created from the physical functions of the PCI cards that are mounted on SLOT4 or higher cannot be assigned to logical domains.
[Example of command output]
# ldm ls-io -l
NAME                       TYPE BUS   DOMAIN  STATUS
----                       ---- ---   ------  ------

...

/SYS/PCI1/SLOT5            PCIE PCIE1 primary OCC [pci@8100/pci@4/pci@0/pci@1/pci@0/pci@0/pci@0/pci@1/pci@0/pci@10/pci@0/pci@1]
    network@0
    network@0,1
...

/SYS/PCI1/SLOT2/IOVNET.PF0 PF   PCIE1 primary
[pci@8100/pci@4/pci@0/pci@1/pci@0/pci@0/pci@0/pci@1/pci@0/pci@10/pci@0/pci@1/network@0]
    maxvfs = 7
...
Workaround This has been modified with SRU 11.2.2.5.0 and patch 150817-03 for Oracle Solaris 10.
Bug 18615814
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description An I/O domain may output the following message, and Oracle Solaris panic may occur if a system board is deleted by executing dynamic reconfiguration of physical partitions (PPAR DR) or a PCIe end point device is dynamically removed from the I/O domain by executing the ldm remove-io command.
panic[cpuX]/thread=XXXXXXXXXXXX: mutex_exit: not owner, lp=XXXXXXXX owner=X thread=XXXXXXXXXXXX
Workaround This has been modified with SRU 11.2.8.4.0.
Execute the svcadm(1M) command on the I/O domain to disable the intrd(1M) service before deleting the system board by executing dynamic reconfiguration of physical partitions (PPAR DR) or before removing the PCIe end point device from the I/O domain.
# svcadm disable intrd
Enable the intrd(1M) service after the process of the ldm remove-io command is completed.
# svcadm enable intrd
Bug 18665751
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description When using XCP 2210, the Dynamic Resource Management (DRM) feature of Oracle VM Server for SPARC does not work.
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
Update the XCP firmware to XCP 2220 or later.
Bug 18747641
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description Core dumps may be produced or wrong calculation results may be obtained or a panic may occur when a program, which performs double-precision floating point instructions after enabling SPARC64 X/SPARC64 X+ processor-specific options and 4-byte boundary alignment (*1) and compiled with Oracle Solaris Studio compiler version 12.3 2013/06/17 or newer, is executed on a SPARC M10 system with Oracle Solaris 11.1 or newer.
*1 The 4-byte boundary alignment is enabled by default when creating 64-bit programs.

In case of 32-bit programs, it is enabled if "-xmemalign=Ns (N=1,2,4,8,16)" or "-fast" is not specified.


[Procedure of checking compiler version]
The "-V" option shows version information. The date is output at the end of version notation. The compiler version that corresponds to this bug is 2013/06/17 or newer.
$ cc -V
cc: Sun C 5.12 SunOS_sparc Patch 148917-06 2013/06/17
$ f95 -V (f90 and f77 are also same.)
f95: Sun Fortran 95 8.6 SunOS_sparc Patch 148517-05 2013/06/17
$ CC -V
CC: Sun C++ 5.12 SunOS_sparc Patch 148506-11 2013/06/17
Workaround This has been modified with SRU 11.2.4.6.0.
Recompile the program with the following "-xarch" flag.
-xarch=sparcima
Bug 19074260
Model SPARC M10-4S
Description The following messages may be output in the log of ldoms/ldmd services (/var/svc/log/ldomsldmd:default.log), and the communication between the ldmd daemon and the XSCF may be disconnected during or after physical partition dynamic reconfiguration (PPAR DR).
[Example of message]
Sep 18 13:31:37 warning: Device busy: open_ldc_channel: Open of/devices/virtual-devices@100/channel-devices@200/virtual-channel@3:spds failed
After that time, processes which need to communicate with XSCF such as PPAR DR or ldm list-spconfig command fail.
Workaround This has been modified with SRU 11.2.8.4.0.
[How to restore]
Execute Oracle Solaris svcadm(1M) command to restart the ldoms/ldmd services.
# svcadm restart ldmd
Bug 19310540
Model SPARC M10-4S
Description If the addboard(8) command is executed in the "factory-default" configuration, CPU cores may not be assigned to the control domain.
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
Add the CPU cores or threads which were not added, using the ldm add-core command or the ldm add-vcpu command.
Bug 19310550
Model SPARC M12-2S, SPARC M10-4S
Description On a physical partition, to which 8 or more system boards have been assigned, when collecting dump files of the hypervisor which is executed as the ldoms/ldmd service is started, the following console message is output by the ldoms/ldmd service, and it may fall back to maintenance mode.
[Example of message]
Feb 28 16:19:39 svc.startd[11]: ldoms/ldmd:default failed:
transitioned to maintenance (see 'svcs -xv' for details)
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
[How to restore]
Use the following process to change the timeout value of starting the ldoms/ldmd service to 600.
# svccfg -s ldmd listprop
:
start/timeout_seconds count 180
:
# svccfg -s ldmd setprop start/timeout_seconds=600
# svccfg -s ldmd listprop
:
start/timeout_seconds count 600
:
# svcadm refresh ldmd
# svcadm restart ldmd
Bug 19358400
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If the root complex is dynamically added/deleted, the PCIe end point device configuration information shown by the showpparinfo(8) command will not reflect the PCIe end point device under the added/deleted root complex.
Workaround This has been modified with SRU 11.2.9.5.0.
[How to restore]
By restarting the logical domain that added/deleted the root complex dynamically, the showpparinfo(8) command displays the correct configuration information.
Bug 19424242
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description On a system to which Oracle VM Server for SPARC 3.1.0.1 or later is applied, the following event may occur: If all CPUs or memory in an I/O domain are degraded due to a CPU or memory failure, the ldmd service abnormally terminates, and, as a result, the ldm(1M) command terminates with an error.
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
[How to restore]
Replace the faulty CPU or memory.
If you want to start Oracle Solaris while leaving the faulty CPU or memory installed, perform the following procedure on the XSCF:
1. Power off the physical partition (PPAR) by the poweroff(8) command.

2. Execute the setdomainconfig(8) command to place the PPAR in the factory-default state.

XSCF> setdomainconfig -p ppar_id -c default
3. Execute the poweron(8) command to activate the PPAR.
Oracle Solaris reboots in a configuration that includes only the control domain (factorydefault).
Bug 19424359
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If the domain configuration is restored in the degraded configuration, both of the following settings are reset to their default values: the setting specifying whether to enable/disable hypervisor dump collection and the setting specifying whether to enable/disable automatic reboot during hypervisor dump collection.
[Default values]
Hypervisor dump collection: Enabled
Automatic reboot during hypervisor dump collection: Disabled
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
[How to restore]
After executing the Oracle VM Server for SPARC ldm(1M) command to change the hypervisor dump setting, save the domain configuration information.
# ldm set-hvdump hvdump=XXXX hvdump-reboot=YYYY
# ldm add-spconfig ZZZZ
After replacing the faulty component, execute the setdomainconfig(8) command to initiate a reboot with the original domain configuration.
Bug 19513561
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description The ldmd daemon of Oracle VM Server for SPARC may repeat core dump if suspend processing of the appropriate domain fails during live migration.
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
[How to restore]
Restart the physical partition according to the following steps.
1. Execute the poweroff(8) command to power off the physical partition (PPAR).

2. Execute the poweron(8) command to restart PPAR.
Bug 19680186
19454809
Model SPARC M12-2S, SPARC M10-4S
Description If Oracle Solaris 11.2 and later is running and the system board is deleted by dynamic reconfiguration of physical partitions (PPAR DR), Oracle Solaris may panic.
Workaround This has been modified with SRU 11.2.10.5.0.
Add the following setting to /etc/system of all logical domains, and restart Oracle Solaris:
set lgrp_topo_levels=1
Be sure to delete set lgrp_topo_levels=1 in /etc/system before applying SRU 11.2.10.5.0 or later.
Bug 19728345
Model SPARC M10-4S
Description The physical partition dynamic reconfiguration (PPAR DR) fails if the ldoms/ldmd services are restarted because of Oracle Solaris panic and the like during PPAR DR.
Workaround This has been modified with SRU 11.2.8.4.0 and Oracle VM Server for SPARC 3.2 for Oracle Solaris 10.
[How to restore]
Hypervisor abort may be caused from the operation of adding/removing memory to/from PPAR DR or a logical domain after the ldoms/ldmd services are recovered. Therefore, execute the poweroff(8) command of XSCF firmware to power off the physical partition (PPAR), then execute the poweron(8) command to power on the PPAR.
Bug 19913088
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If a root complex with PCI expansion unit connected is added dynamically to the logical domain with the ldm add-io command, the logical domain may output the following message, causing Oracle Solaris to panic.
panic[cpuX]/thread=XXXXXXXXXXXX: bad stack overflow at TL 1
Workaround This has been modified with SRU 11.2.10.5.0.
Before adding the root complex to the logical domain dynamically, add the following setting to /etc/system, and then restart Oracle Solaris.
set default_stksize = 0xa000
Bug 20061005
19200041
Model SPARC M10-4S
Description If you use the ipadm(1M) command or the ifconfig(1M) command on the guest domain that has the physical device after you delete the system board dynamically with the deleteboard(8) command, the guest domain may output the following message, causing Oracle Solaris to panic.
panic[cpuXX]/thread=XXXXXXXXXXXX:
assertion failed: obj->afo_corep == NULL, file: ../../common/os/numaio.c,
line: 724
Workaround This has been modified with SRU 11.2.10.5.0.
If you delete the system board dynamically with the deleteboard(8) command, execute the following command before you execute the ipadm(1M) command or the ifconfig(1M) command on the guest domain.
# modunload -i 0
Bug 20458698
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description No response time from the migration source domain may become long because a different process from the original live migration is performed during live migration. Network services and the like operating on the migration source domain may time out because of no response.
This case occurs if the migration source domain meets both of the following conditions.
- The difference between the maximum RA of the migration source domain (actual address) and its minimum RA cannot be divided by 64 MB

- The remainder is 32 MB or less when the difference between the maximum RA of the migration source domain and its minimum RA is divided by 64 MB

The maximum RA and the minimum RA of the domain can be checked with the following command.
[Example]
# ldm list-domain -o memory domain-name
NAME
domain-name
MEMORY
RA         PA              SIZE
0x10000000  0x7b0fc0000000 1G
    minimum RA
0x400800000 0x7f01a0800000 11G
        (a)                (b)
The maximum RA is the sum of (a) + (b), which will be 0x6c0800000.
0x400800000 + 0x2c0000000(11G) = 0x6c0800000
The difference between the maximum RA and the minimum RA is 27400 MB.
0x6c0800000 - 0x10000000 = 0x6b0800000 = 27400 MB
Therefore, the remainder is 8 MB in this example.
27400 MB / 64 MB = 428 and the remainder is 8 MB
Workaround This has been modified with SRU 11.2.11.5.0.
Bug 20878144
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description For Oracle Solaris 11.2 SRU 11.2.8.4.0 or later, "OS Started. No state support" is displayed by the showdomainstatus(8) command or in the event log when Oracle Solaris starts. This message indicates that the status of a logical domain has changed to Oracle Solaris.
The following is an example of the message.
XSCF> showlogs event
Date Message
--- Omitted ---

Mar 27 15:55:31 ** Event: SCF:PPARID 0 GID 00000000 state change (OpenBoot Running)
Mar 27 15:55:32 ** Event: SCF:PPARID 0 GID 00000000 state change (OpenBoot Primary Boot Loader)
Mar 27 15:55:33 ** Event: SCF:PPARID 0 GID 00000000 state change (OpenBoot Running OS Boot)
Mar 27 15:55:35 ** Event: SCF:PPARID 0 GID 00000000 state change (OS Started. No state support)
Mar 27 15:55:36 ** Event: SCF:PPARID 0 GID 00000000 state change (OS Started. No state support)
Mar 27 15:56:42 ** Event: SCF:PPARID 0 GID 00000000 state change (Solaris booting)
Mar 27 15:57:37 ** Event: SCF:PPARID 0 GID 00000000 state change (Solaris booting)
Mar 27 15:57:37 ** Event: SCF:PPARID 0 GID 00000000 state change (Solaris running)
XSCF> showdomainstatus -p 0
2015-MM-DD hh:mm:ss
Logical Domain Name Status
primary OS Started. No state support.
Workaround This has been modified with SRU 11.2.11.5.0.
Ignore this message since it does not affect the system operation.
Bug 20974426
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description In an environment with Oracle VM Server for SPARC 3.2 applied to the control domain and configuration information already saved in the XSCF, if the SPARC M10 system chassis or physical partition (PPAR) is stopped or started, the SPARC M10 system chassis or PPAR may not be able to start with the saved configuration information.
This problem occurs when the configuration information is saved by any of the following means:
- ldm add-spconfig -r command

- Automatic recovery using automatic recovery policy 3 of Oracle VM Server for SPARC ldmd daemon (automatic saving of configuration information)

You can check the automatic recovery policy of the ldmd daemon with the following command.
The default for the automatic recovery policy is 1 (display warning messages in log files)
[Example]
# svccfg -s ldmd listprop ldmd/autorecovery_policy
ldmd/autorecovery_policy integer 3
Workaround This has been modified with SRU 11.2.11.5.0.
[How to restore]
- If the ldm add-spconfig -r command was executed, delete the saved configuration information, and overwrite it by saving the current configuration.

[Example]
# ldm remove-spconfig CONF-A
# ldm add-spconfig CONF-A

- If the automatic recovery policy is set to 3, change the automatic recovery policy to 1 by performing the following procedure.

[Example]
# svccfg -s ldmd setprop ldmd/autorecovery_policy=1
# svcadm refresh ldmd

If the SPARC M10 system chassis or PPAR cannot start with the saved configuration information, start the system in the factory-default configuration, and then restore the configuration information already saved in the XML file.
Bug 21106074
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If any of aes-128-ccm, aes-192-ccm, and aes-256-ccm is specified for the encryption algorithm, accessing the ZFS data set with encryption enabled may cause a system panic.
The default encryption algorithm is aes-128-ccm. If ZFS is encrypted with no encryption algorithm specified, aes-128-ccm is assumed specified.
[Panic message example]
panic[cpu34]/thread=2a1053d9c20: bad floating point trap at TL 1
%tl %tpc %tnpc %tstate %tt
1 00000000123eabc0 00000000123eabc4 8880001600 077
%gl: 00 %ccr: 88 %asi: 80 %cwp: 0 %pstate: 16
(Omitted)
Workaround This has been modified with SRU 11.2.12.5.0.
Add the following statements to the /etc/system file, and reboot the system.
set auxv_cap_exclude_hw1=0x10000
set auxv_cap32_exclude_hw1=0x10000
Bug 21306352
Model SPARC M12-2S, SPARC M10-4S
Description The physical partition dynamic reconfiguration (PPAR DR) feature may fail if used to delete a system board in an environment containing a root domain (not a control domain) running Oracle Solaris 11.2 SRU 11.2.9.5.0 or later.
[Example]
XSCF> deleteboard -y -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]
0.end

Unconfigure preparation of PSB has completed.
Start unconfiguring PSB from PPAR. [7200sec]
0..... 30..... 60..... 90.....-

end
Timeout detected during communicate with Logical Domains Manager.
XSCF>
Workaround You can avoid this problem by deleting the PCIe bus on the target system board from the domain before the PPAR DR feature deletes the system board.
[Example]
primary# ldm remove-io PCIE8 domainX
:
primary# ldm remove-io PCIE15 domainY
XSCF> deleteboard -y -c disconnect -m unbind=resource 01-0
[How to restore]
After deleting the PCIe bus on the target system board from the domain, re-execute the deleteboard command on the XSCF.
[Example]
primary# ldm remove-io PCIE8 domainX
:
primary# ldm remove-io PCIE15 domainY
XSCF> deleteboard -y -c disconnect -m unbind=resource 01-0