Skip to main content

Problems Resolved in Oracle Solaris 11.2


Problems Resolved in Oracle Solaris 11.2
The following table lists the problems resolved in Oracle Solaris 11.2. You might encounter them in supported releases earlier than Oracle Solaris 11.2.
Table 3-53  Problems Resolved in Oracle Solaris 11.2
Bug 15812880
Model SPARC M12-2S, SPARC M10-4S
Description If you try to access, via telnet or ssh, a domain where 8000 GB (about 7.8 TB) or greater memory is mounted, the following message appears on the destination control domain console and the access fails.
- For ssh

error: /dev/ptmx: Not enough space
error: session_pty_req: session 0 alloc failed
- For telnet

telnetd: open /dev/ptmx: Not enough space
Workaround This has been modified with Oracle Solaris 11.1 SRU 3.5.1 and patch 148888-04 for Oracle Solaris 10.
Execute the following command to change ptmx_ptymax:
[Example]
# echo "ptms_ptymax/Z 0x400000" | mdb -kw
ptms_ptymax: 0 = 0x400000
Bug 15822113
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If ldm add-vcpu and ldm remove-vcpu are repeatedly executed in a shell script, the process that is being run may cause a core dump and abnormally terminate.
Workaround This has been modified with SRU 11.1.7.5.0.
If this defect occurs because SRU is not applied, execute the command again. 
In addition, when executing ldm remove-vcpu, execute it in such a condition that the process load is low.
Bug 15823255
Model SPARC M12-2S, SPARC M10-4S
Description An Oracle Solaris panic may occur if the CPU allocation is changed using Oracle Solaris psradm(1M) command or psrset(1M) command or the configuration of a virtual CPU is dynamically changed using the Oracle VM Server for SPARC ldm command under the environment that meets the following two conditions.
- The environment where the physical partition (PPAR) is composed of two or more SPARC M10-4S chassis.

- The environment where the following lgroup exists when the lgrpinfo command is executed on the control domain or logical domain.

Among the lgroups that are displayed as "lgroup XX (intermediate):",
only one number is displayed before (CPU) of the "Lgroup resources:" field.
This number is not displayed before (memory).
[Example]
# /usr/bin/lgrpinfo
...

lgroup 12 (intermediate):
        Children: 10, Parent: 0
        CPUs: 0 1
        Memory: installed 520M, allocated 494M, free 26M
        Lgroup resources: 1 (CPU); 10 11 (memory)
        Latency: 21
...
Workaround This has been modified with Oracle Solaris 11.1 SRU 5.5 and patch 150400-01 for Oracle Solaris 10.
Add the following line to /etc/system and restart Oracle Solaris:
set lgrp_topo_levels=2
Bug 15825208
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description In SPARC M10 Systems, the scp(1), sftp(1), and ssh(1) commands of Oracle Solaris may generate an error or installation of Oracle RAC may fail.
Workaround This has been modified with Oracle Solaris 11.1 SRU 1.4.
See "An Error Occurs in Executing the scp(1), sftp(1), or ssh(1) Command of Oracle Solaris or Installation of Oracle RAC Fails (CR:15825208)."
Bug 15826052
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description You cannot use the PCI hot plug (PHP) function to add a Quad Gigabit Ethernet card (SE1X7GQ2F) to a PCI Express slot of a PCI expansion unit.
Workaround This has been modified with XCP 2050 and SRU 11.1.6.4.0.
For XCP 2050, see the description of RTIF2-130528-001.
If XCP and SRU are not applied, stop the logical domain to which you want add the PCI card before adding it.
Bug 15840018
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description After the firmware is updated to XCP 2031 or later on the SPARC M10 or to XCP 3021 or later on the SPARC M12, the following message appears at Oracle Solaris startup.
NOTICE: skipping unsupported token: fjorclnum
Workaround This has been modified with SRU 11.1.4.5.0 and Oracle Solaris 10 patch 148888-03.
Ignoring this message does not have an impact on the system.
Bug 15851224
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description When starting the I/O domain, the following message may be output and the panic may persist.
recursive rw_enter, lp=XXXXXXXX wwwh=XXXXXXXX thread=XXXXXXXX
Workaround This has been modified with SRU 11.1.12.5.0 and patch 150840-01 for Oracle Solaris 10.
Add the following to /etc/system of the I/O domain and restart Oracle Solaris:
forceload:  drv/vpci

Note that if starting of the I/O domain is disabled, you must change to inactive state the guest domain to which the virtual disk (vdisk), which is a target of the virtual disk service (vds) of the I/O domain, is assigned, with the ldm stop-domain and ldm unbind-domain commands. This will enable starting of the I/O domain. After starting the I/O domain, set the above.
Bug 15851441
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description When the memory fails and is degraded, the following message may be output at Oracle Solaris startup, and the panic may persist:
tilelet_assign_fini_cb(): tile 0xX in memgrp X was unused

This may also occur when you set the mirror mode for the memory after saving the logical domain setting with ldm add-spconfig.
Workaround This has been modified with SRU 11.1.11.4.0.
See "If Oracle Solaris is Activated During Memory Degradation, a Panic May Occur (CR:15851441)."
Bug 15858713
16769782
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If memory fault occurs and all memories allocated to guest domains are degraded, Oracle VM Server for SPARC ldmd daemon generates a core dump and terminates abnormally, and the ldm(1M) command ends with an error.
Workaround This has been modified with SRU 11.1.10.5.0. On Oracle Solaris 10, it has been modified with Oracle VM Server for SPARC 3.1.
If this defect occurs because SRU is not applied, replace the memory having an error.
If you want to start Oracle Solaris while the memory having an error remains mounted, start it using the following procedure from XSCF.
1. Execute the poweroff(8) command to power off the physical partition (PPAR).

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 restart PPAR.

Oracle Solaris reboots in a configuration that includes only the control domain (factorydefault).
Bug 15887244
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description When you start the SunVTS 7.0 ps14 and ps15 tests in a SPARC M10 system, they may terminate with an error.
Workaround This has been modified with Oracle Solaris 11.1 SRU 4.6 and patch 149395-02 for Oracle Solaris 10.
There is no workaround other than applying the modification.
Bug 16238762
Model SPARC M12-2S, SPARC M10-4S
Description If a system board is added to the PPAR by dynamic reconfiguration of physical partitions after booting the Oracle Solaris in the control domain with the factory-default configuration, or if CPUs on the system board are added to a domain by the ldm add-vcpu command after the system board is added by dynamic reconfiguration of physical partitions, the system will panic with the following message:
panic[cpuX]/thread=XXXXXXXXXXX: mpo_cpu_add: Cannot read MD
Workaround This has been modified with SRU 11.1.7.5.0 and patch 150400-12 for Oracle Solaris 10. Meanwhile, when this fix has been applied to the system, if a system board is added to the system by means of the dynamic reconfiguration of physical partitions, the following message may be output but as it has no effect on the system, ignore it.
WARNING: mpo_cpu_add: defaulting to lgroup x for CPU x
For the control domain or the domain whose CPU is added by the ldm add-vcpu command, add the following line to the domain's /etc/system file and reboot the Oracle Solaris.
set suspend_count = 1
Bug 16292272
Model SPARC M12-2S, SPARC M10-4S
Description If you configure many guest domains in a system where 16 BBs compose one physical partition (PPAR), it takes time to perform binding for the guest domains.
The time taken by new binding processing for each guest domain is approximately (number of guest domains that are already bound + 1) x 6 + 10 seconds.
Therefore, if there are no domains for which binding has been performed, and binding is performed for the guest domains one by one through the ldm bind-domain command, the required time is obtained by summing the times taken to perform binding for all of them.
Workaround This has been improved with SRU 11.1.16.4.0 and patch 150011-03 for Oracle Solaris 10.
We recommend that you do not configure the system with a single PPAR but divide it into multiple PPARs and then configure guest domains in each PPAR.
Using the recommended configuration described above, not only mitigates the described phenomenon but also helps improve fault tolerance.
If a virtual network switch (vsw) is configured, you can reduce the time for binding by half by setting inter-vnet-link to off. For notes when inter-vnet-link is set to off, see the Oracle VM Server for SPARC Administration Guide or Oracle VM Server for SPARC Release Notes.
Bug 17510986
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If the guest domain is Oracle Solaris 11.1 and SRU 11.1.9.5.1 or later is applied to it, or if the guest domain is Oracle Solaris 10 and 150400-01 or later is applied to it, system panic may occur during live migration or dynamic reconfiguration of physical partitions.
Workaround This has been modified with SRU 11.1.14.5.0 and patch 150400-07 for Oracle Solaris 10.
Bug 17627526
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description Messages like the following are output into the console at the time of starting Oracle Solaris and the ldoms/ldmd service is put into maintenance mode.
[Example of message]
Jan 20 16:01:37 svc.startd[11]: svc:/ldoms/ldmd:default: Method "/opt/SUNWldm/bin/ldmd_start" failed with exit status 96.
Jan 20 16:01:38 svc.startd[11]: ldoms/ldmd:default misconfigured: transitioned to maintenance (see 'svcs -xv' for details)
At this time, messages like the following will be registered to the log file (/var/svc/log/ldomsldmd:
[Example of message]
[ Jan 21 20:08:55 Executing start method ("/opt/SUNWldm/bin/ldmd_start"). ]
ldmd cannot communicate with the hypervisor as the required device does not exist:
/devices/virtual-devices@100/channel-devices@200/virtual-channel@0:hvctl
Workaround This has been modified with SRU 11.1.19.6.0 and patch 150840-04 for Oracle Solaris 10.
[How to restore]
After confirming that the device file in question exists, restore the ldoms/ldmd service using the svcadm(1M) command of Oracle Solaris.
# ls -l
/devices/virtual-devices@100/channel-devices@200/virtual-channel@0:hvctl
crw-------   1 root     sys      148, 2048 Jan  21 20:08 /devices/virtual-devices@100/channel-devices@200/virtual-channel@0:hvctl
# svcadm clear ldmd
Bug 17709858
Model SPARC M12-2S, SPARC M10-4S
Description If the system board is deleted by dynamic reconfiguration of the physical partition, any of the following phenomena may occur.
- The following message is output from the logical domain, and Oracle Solaris enters the panic state. Fatal error has occurred in: PCIe fabric.(0xxx)(0xxx)

- There is a hang during the device resume processing of a logical domain, and deleteboard(8) ends with an error due to a timeout.
Workaround This has been modified with SRU 11.1.15.4.0.
Bug 17777004
Model SPARC M10-4S
Description If the logical domain has degraded memory because of a memory failure, when dynamic reconfiguration of the physical partition is executed with the deleteboard(8) command, Oracle Solaris on the domain with the degraded memory may enter the panic state.
Workaround This has been modified with SRU 11.1.17.5.0 and patch 150817-02 for Oracle Solaris 10.
Before executing dynamic reconfiguration of the physical partition, check whether memory degradation has occurred from the control domain. If memory degradation has occurred, delete the associated memory area in advance.
[How to check]
Execute "ldm list-devices -a -S memory". If the resulting STATUS column displays "fail", the memory area has been degraded by SIZE from PA (physical address) displayed on the same line.
[How to delete a memory area]
Execute "ldm remove-memory <total_value_of_above_SIZEs>‌ <‌domain_name>‌".
Bug 18055846
Model SPARC M12-2, SPARC M12-2S, SPARC M10-4, SPARC M10-4S
Description When adding a PCI expansion unit by using PCI hot plug (PHP), Oracle Solaris may panic by producing the following message.
panic[cpuX]/thread=XXXXXXXXXX: Fatal error has occurred in: PCIe fabric.(0xX)(0xXX)
Workaround This has been modified with Oracle Solaris 11.2, and patch 150400-18 for Oracle Solaris 10.
Bug 18112775
Model SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description When Oracle VM Server for SPARC 3.1.0.1/SRU 11.1.14.5.0 is installed on the control domain, if dynamically assigned memory is reduced from a guest domain, which is running Oracle Solaris 10, with the ldm set-memory or the ldm remove-memory command, Oracle VM Server for SPARC ldmd daemon on the control domain may produce a core dump and it may be restarted.
Workaround This has been modified with SRU 11.1.17.5.0 and patch 150817-02 for Oracle Solaris 10.
Check the memory size that has been assigned to a logical domain with the ldm list-domain command and reduce it to the required size in several steps, using the ldm remove-memory command. It is recommended to reduce memory size by less than 256 MB at one time, but memory may also be reduced by a larger amount at one time.
Meanwhile, the reduction may fail due to the occurrence of the same symptom while performing this action. In such a case, reduce the memory into smaller units.
Bug 18009532
Model SPARC M12-2S, SPARC M10-4S
Description If dynamic reconfiguration is performed on a physical partition which has a dual channel 10Gbps FCoE card (SP1X7FAR2F/SP1X7FAS2F/7101673 (7101674)+7101677 (7101678)) mounted on it, system panic may occur due to the suspend processing of the inactivated qlcnic driver.
Workaround This has been modified with SRU 11.1.19.6.0 and patch 149167-03 for Oracle Solaris 10.
Before dynamically reconfiguring a physical partition on which this specific card is mounted, activate any unplumbed qlcnic interface with the ifconfig interface_name‌ plumb command.
Bug -
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description If Java VM is started on a SPARC M12 system or on a SPARC M10 system with the SPARC64 X+ processor, the following message may be output along with the production of a core dump.
[Example of message]
# A fatal error has been detected by the Java Runtime Environment:
#
#  Internal Error (output.cpp:1576), pid=1310, tid=91 #  guarantee((int)(blk_starts[i+1] - blk_starts[i]) >= (current_offset
- blk_offset)) failed: shouldn't increase block size # <...> # Abort (core dumped)


Moreover, due to the production of the core dump by Java VM, the following events may occur:
1. On a SPARC M12 system or on a SPARC M10 system with the SPARC64 X+ processor, the [Add Asset] operation on Oracle Solaris fails.

2. On a SPARC M12 system or on a SPARC M10 system with the SPARC64 X+ processor, the [Install Server] operation for Oracle Solaris installation fails.

3. On a SPARC M12 system or on a SPARC M10 system with the SPARC64 X+ processor, Enterprise Controller and Proxy Controller (Ops Center management server) installation on the domains fail.
Workaround This has been modified with SRU 11.1.17.5.0.
Bug -
Model SPARC M10-4S
Description If dynamic reconfiguration is performed on a physical partition which has a Dual 10Gbps FCoE card (SE0X7EC12F/SE0X7EF12F/SG-PCIEFCOE2-Q-TA (SG-XPCIEFCOE2-Q-TA, 7105382)/SGPCIEFCOE2-Q-SR (SG-XPCIEFCOE2-Q-SR,7105381)) mounted on it, system panic may occur due to the processing of the qlge driver.
Workaround This has been modified with SRU 11.1.8.4.0 and patch 145648-04 for Oracle Solaris 10.
There is no workaround other than applying the modification.
If this specific card is mounted on a system, conduct any reconfiguration in deactivated state instead of dynamic reconfiguration.
Bug -
Model SPARC M12-1, SPARC M12-2, SPARC M12-2S, SPARC M10-1, SPARC M10-4, SPARC M10-4S
Description After the firmware is updated to XCP 2250 or later on the SPARC M10 with the SPARC64 X+ processor or to XCP 3021 or later on the SPARC M12, the following message appears at Oracle Solaris startup.
[SPARC M10]
  NOTICE: skipping unsupported token: fjathplus
  NOTICE: skipping unsupported token: vis3b
[SPARC M12]
  NOTICE: skipping unsupported token: fjathplus
  NOTICE: skipping unsupported token: vis3b
  NOTICE: skipping unsupported token: fjathplus2
  NOTICE: skipping unsupported token: vis3c
  NOTICE: skipping unsupported token: sparc5b
Workaround This has been modified with patch 148888-03 for Oracle Solaris 10.
Ignoring this message does not have an impact on the system.
An Error Occurs in Executing the scp(1), sftp(1), or ssh(1) Command of Oracle Solaris or Installation of Oracle RAC Fails (CR:15825208)
[Workaround]
Using the following procedure, change the setting so that the AES_CTR, AES_CBC_ MAC, and AES_CFB128 algorithms are not used with the assist function of the encryption unit.
  1. If you use it with the client function (scp(1), sftp(1), ssh(1), etc.):
    1. Add the contents of the setting to the target file as one line. A space is necessary only between "Cipher" and "aes128-cbc."
    - Target file
    Settings for entire system: /etc/ssh/ssh_config
    Settings per user: $HOME/.ssh/ssh_config
    - Setting contents
Ciphers aes128-cbc,aes192-cbc,aes256-cbc,3des-bc,arcfour128,arcfour256,arcfour,blowfish-cbc
  1. If you use it with the server function (such as Oracle Solaris sshd daemon)
    1. Add the contents of the setting to the target file as one line.
    - Target file
    /etc/ssh/ssh_config
    - Setting contents (recommended)
Ciphers 3des-cbc,arcfour128,arcfour256,arcfour,blowfish-cbc
  1. 2. Restart the service with the following command:
# svcadm restart svc:/network/ssh:default
  1. If the problem cannot be resolved with the above setting in either the server function or client function, whichever is used
    1. Add the following, in place of the above content of the setting:
UseOpenSSLEngine no
If Oracle Solaris is Activated During Memory Degradation, a Panic May Occur (CR:15851441)
[Workaround]
If a panic occurs due to a memory failure, replace the failed memory.
If a panic occurs due to the memory mirror mode setting, start the system from the XSCF by using the following procedure.
Also, if a panic occurs due to a memory failure, starting the system from the XSCF by using the procedure shown below may prevent the problem, but this method is not always reliable. If a panic recurs even after the system is started by using the following procedure, replace the failed memory.
  1. Execute the poweroff(8) command to power off the physical partition (PPAR).
  2. Execute the setdomainconfig(8) command to place the PPAR in the factory-default state.
XSCF> setdomainconfig -p ppar_id -c default
  1. Execute the poweron(8) command to activate the PPAR.
Oracle Solaris reboots in a configuration that includes only the control domain (factorydefault).
If you set the mirror mode for the memory, configure a logical domain after setting the mirror mode in the factory-default condition. Then, save the settings with ldm add-spconfig.
If the settings were saved with the mirror mode not set, do not specify them in the ldm set-spconfig command or the XSCF setdomainconfig(8) command for an environment using the mirror mode.