The following hardware–related issue and bugs apply to the Solaris 10 release.
The ixgbe driver has 8 RX queues by default. The multiple RX queues result in reduced driver performance in the promiscuous mode when compared to a single RX queue in the Solaris 10 10/08 and Solaris 10 5/09 releases. Multiple RX queues create kernel contentions. These kernel contentions might contribute to poor ixgbe performance in the Solaris 10 10/09 release.
Workaround: Set the ixgbe RX and TX queue values to 1. Perform the following steps:
Edit the /kernel/drv/ixgbe.conf configuration file.
rx_queue_number = 1; tx_queue_number = 1; |
Reboot the system.
After reboot the ZFS root pool comes up in degraded mode. A mirrored root pool that might be impacted by this issue will contain disk devices that use different device drivers and the second driver isn't loaded during the boot process.
A symptom of this issue is when the system is rebooted, the pool state is DEGRADED because both device drivers are not loaded at the same time. A mirrored root pool that might be impacted by this issue will have zpool status output similar to the following:
state: DEGRADED status: One or more devices could not be opened. Sufficient replicas exist for the pool to continue functioning in a degraded state. |
Workaround: Perform the following steps:
Identify the device drivers to be force loaded. For example, use the prtconf command with syntax similar to the following:
# prtconf -aD /dev/rdsk/c1t0d0 SUNW,Sun-Fire-280R (driver name: rootnex) pci, instance #1 (driver name: pcisch) SUNW,qlc, instance #4 (driver name: qlc) fp, instance #5 (driver name: fp) ssd, instance #5 (driver name: ssd) |
Edit the /etc/system file and provide a forceload entry for each driver identified in step 1. For example:
forceload: drv/qlc forceload: drv/ssd forceload: drv/fp |
Certain x86 systems fail to plumb the fourth 5709 interface with 1 Gbyte of RAM. The system might panic or the following error message is displayed:
ifconfig: plumb: bnx<X>: Invalid argument |
This bug affects Dell PowerEdge R610 and R710 servers.
Workaround: Perform the following steps:
Use more than 1 Gbyte of RAM.
Create a hostname.bnxX for each interface.
Reboot your system.
Download and install the latest BNX driver 4.6.2.
Modify the bnx.conf file to use the default setting for tx_descriptor_count.
Upgrading the ConnectX firmware to version 2.6.000 or higher on a running system might cause problems in some HCAs and x64 platforms. This issue affects only Mellanox-branded HCAs. Sun-branded PCIe HCAs, EMs, NEMs, and SPARC platforms are not affected.
You might be unable to boot the system or the systrem might hang during boot. ibd (IPoverIB) instance numbers might change and prevent the system from booting and plumb ibd devices.
Workaround 1: Before rebooting the system after updating the firmware using cxflash, remove the ibd<x> instances from the /etc/path_to_inst file and /dev directory. Perform the following steps:
Log in as the root user. The device tree information might look like this:
# ls -R /devices | grep 15b3 /devices/pci@0,0/pci10de,5d@e/pci15b3,634a@0 /devices/pci@0,0/pci10de,5d@e/pci15b3,634a@0/ibport@1,ffff,ipib /devices/pci@0,0/pci10de,5d@e/pci15b3,634a@0/ibport@1,ffff,ipib:ibd0 /devices/pci@0,0/pci10de,5d@e/pci15b3,634a@0/ibport@2,ffff,ipib /devices/pci@0,0/pci10de,5d@e/pci15b3,634a@0/ibport@2,ffff,ipib:ibd1 /devices/pci@0,0/pci10de,5d@e/pci15b3,634a@0:devctl |
Device specifics vary depending on the Mellanox-branded card installed.
Edit the path_to_inst file. Perform the following steps:
Make a backup copy of the existing path_to_inst file.
# cp /etc/path_to_inst /etc/path_to_inst.backup |
Open the path_to_inst file:
# vi /etc/path_to_inst |
Search for lines that contain ibd and hermon, and delete those lines.
Save the changes and close the file.
Remove the entries in the /dev directory:
rm /dev/ibd?* |
Reboot the system. The system boots normally, and the corresponding device tree looks like this:
# ls -R /devices | grep 15b3 /devices/pci@0,0/pci10de,5d@e/pci15b3,5@0 /devices/pci@0,0/pci10de,5d@e/pci15b3,5@0/ibport@1,ffff,ipib /devices/pci@0,0/pci10de,5d@e/pci15b3,5@0/ibport@1,ffff,ipib:ibd0 /devices/pci@0,0/pci10de,5d@e/pci15b3,5@0/ibport@2,ffff,ipib /devices/pci@0,0/pci10de,5d@e/pci15b3,5@0/ibport@2,ffff,ipib:ibd1 /devices/pci@0,0/pci10de,5d@e/pci15b3,5@0:devctl |
Workaround 2: If you reboot the system before updating the path_to_inst file and /dev directory, the system might hang. In such a case, perform the following steps:
Power off the system and remove the HCA from the bus.
Reboot the system without the HCA installed.
When system is back on, follow steps 2 through 3 listed in Workaround 1.
Power off the system.
Re-intsall the HCA. Reboot the system.
If the system reboots, but ibd interfaces do not automatically plumb, the /etc/hostname.ib<?> files might not be appropriate for the current device configuration. Manually rename the files to reflect the correct configuration.
Some IBM systems either hang or panic after install. The following panic message is displayed:
Unrecoverable Machine-Check Exception |
Workaround 1: Add the following line to the /etc/system file after installation to avoid system hang:
set pcie:pcie_aer_ce_mask=0x11C1 |
Workaround 2: If the system panics during install or during boot after the install, then use this workaround. Boot the system, with -kd option and set it in the kmdb debugger:
::bp pcie`pcie_initchild :c pcie_aer_ce_mask/W 11C1 :d all :c |
When the system comes up append the following line to the /etc/system file to make the changes permanent:
set pcie:pcie_aer_ce_mask=0x11C1 |
Systems with the Intel Xeon Processor MP 7400 series running the Solaris 10 10/09 OS might experience reduced performance and increased power consumption under light utilization. This problem might occur when CPUs do not quiesce, preventing power management while idle. No error message is displayed.
Workaround: Add the following line to the /etc/system file and reboot the system:
set idle_cpu_prefer_mwait=0 |
The standard fwflash command needs a plug-in to support firmware update on the ConnectX HCA flash device. This plug-in is not yet available in the Solaris 10 10/09 release.
Workaround: A standalone version of the utility is available to enable Solaris 10 users to update their firmware. Perform the following steps:
Download the utility and man page from the Customer Download Center at http://www.sun.com/download/index.jsp?tab=2. On this page search for Solaris InfiniBand (IB) Updates 3. Follow the instructions for downloading the package and untar the files. Do not install the package since you need to extract only some of the files.
Rename the original .tar file:
# mv ib_updates_3_s10u6.i386.tar ib_updates_3_s10u6.i386.tar-orig |
Unzip the compressed .tar file:
#gunzip ib_updates_3_s10u6.i386.tgz |
Extract the cxflash utility from the .tar file:
# tar -xvf ib_updates_3_s10u6.i386.tar usr/sbin/cxflash |
This will extract the relative path below your working directory:
# ls -lR usr usr: total 2 drwxrwxrwx 2 staff staff 512 Mar 9 09:20 sbin usr/sbin: total 88 -rwxr-xr-x 1 staff staff 44976 Jan 26 13:39 cxflash |
Copy the cxflash file to /usr/sbin. This utility can be used to flash firmware onto hermon and ConnectX HCAs.
# ls -l firmware total 1952 -rw-r--r-- 1 staff staff 487336 Jan 30 18:29 SUN-375-3548_2.5.100.bin -rw-r--r-- 1 staff staff 487436 Jan 30 18:29 SUN-375-3549_2.5.100.bin |
The firmware directory which is created when the files are extracted, contains the firmware files for Sun-branded HCA cards and Express Modules. If you have a different versions of the HCA card, please obtain firmware of minimum 2.5.000 revision from your HCA card vendor.
bge fails to operate normally with the HP NC326i device. No error message is displayed.
Workaround: None.
Performance regression occurs for the Niagara Crypto Provider (NCP) device driver on Sun SPARC® Enterprise T5220 machines with the Solaris 10 10/09 release. No error message is displayed.
Workaround: Add the following line to the /platform/sun4v/kernel/drv/ncp.conf configuration file:
ncp-threads-per-core=6; |
Floppy disks are not automatically mounted when inserted into USB floppy drives. No error message is displayed. The pcfs file system on the floppy disk is not accessible.
Workaround: The file system on the floppy disk can be mounted manually by running the mount command. For example:
mount -F pcfs /vol/dev/dsk/c3t0d0/noname /rmdisk |
The following error message is displayed when a USB storage device is disconnected:
Disconnected device was busy, please reconnect. |
This message is displayed even when the device is unmounted successfully.
Workaround: None. The error message can be safely ignored.
ZFS can potentially allocate kernel memory across all system boards on systems with very large memory configurations. One free system board is required for dynamic memory reconfiguration so that the memory from the board to be dynamically reconfigured can be copied to the free board. The dynamic memory reconfiguration means that you cannot dynamically reconfigure memory on systems with very large memory configurations that have ZFS running. High-end SunFireTM servers can relocate kernel pages so that this issue is avoided. These servers must have kernel page relocation (KPR) enabled for domains with more than 32 cores. No error message is displayed
Workaround: Reduce the amount of kernel memory that ZFS can allocate by setting the zfs_arc_max parameter in the /etc/system file. The following example sets the maximum size to 512 Mbytes.
set zfs_arc_max = 0x20000000 |
The mpathadm show logical-unit subcommand lists the load balancing global configuration value for the Current Load Balance property. However, entries in the csi_vhci.conf that change the load-balance type for a specific product are not reflected in the mpathadm output even when the setting is active.
If the background processes for the registration tool are left running, the Elite3D and Creator3D framebuffers stop power management. This failure reduces the power savings when the system is in a power-managed state. Under certain conditions, sys-suspend might also hang. No error message is displayed. The system might hang during a system suspend or resume operation.
Workaround:
Run the following command approximately 60 seconds after each login:
# pkill -f basicreg.jar # pkill -f swupna.jar |
A new cryptographic framework is provided in Solaris 10 OS. However, versions 1.0 and 1.1 of the Sun Crypto Accelerator 4000 board's software and firmware do not utilize this framework. Consequently, these versions are not supported in the Solaris 10 OS.
The 2.0 release uses the new framework. This release is available as a free upgrade to current Sun Crypto Accelerator 4000 users who plan to use Solaris 10 OS. Because the Sun Crypto Accelerator 4000 is an export-controlled product, you must contact Sun Enterprise Services or your local sales channel to obtain the free upgrade. Additional information is available on the Sun Crypto Accelerator 4000 web page at Sun's products site.
Support for certain USB 2.0 controllers has been disabled because of incompatibilities between these devices and the EHCI driver. The following message is displayed:
Due to recently discovered incompatibilities with this USB controller, USB2.x transfer support has been disabled. This device will continue to function as a USB1.x controller. If you are interested in enabling USB2.x support please refer to the ehci(7D) man page. Please refer to www.sun.com/io for Solaris Ready products and to www.sun.com/bigadmin/hcl for additional compatible USB products. |
For the latest information about USB devices, see http://www.sun.com/io_technologies/USB-Faq.html.
This Solaris release supports both USB 1.1 and USB 2.0 devices. The following table is a summary of USB devices that work in specific configurations. Connection types can either be direct to the computer or through a USB hub. Note that USB 1.1 devices and hubs are low speed or full speed. USB 2.0 devices and hubs are high speed. For details about ports and speeds of operation, see the System Administration Guide: Devices and File Systems.
Table 2–1 USB Devices and Configurations
USB Devices |
Connection Types |
---|---|
USB 2.0 storage devices |
Direct, USB 1.1 hub, USB 2.0 hub |
USB 1.1 devices except audio |
Direct, USB 1.1 hub, USB 2.0 hub |
USB 1.1 audio devices |
Direct, USB 1.1 hub |
USB 2.0 audio devices |
Not supported |
The following list describes limitations with certain drivers and interfaces in this release of Solaris 10 for x86 platforms:
This functionality is turned off for all device types. In the DDI_SUSPEND code in your detach() function, you should return DDI_FAILURE.
This functionality is unavailable to USB devices. Do not create power management components. Write your driver so that pm_raise_power() and pm_lower_power() are called only when power management components are created.
Power management of interactive devices such as removable media is linked with power management of your monitor and the graphics card that drives your monitor. If your screen is active, devices such as the CD-ROM drive and diskette remain at full-power mode. These devices might switch to low-power mode on a system without a monitor. To restore power to the CD or diskette, type volcheck to obtain the latest status from each removable device.
Alternatively, you can disable power management on your system by using the Dtpower GUI. By disabling power management, these devices are constantly at full power.
By default, the kdmconfig program specifies Generic US-English(104-Key) as the keyboard type that is connected to the system. If the system's keyboard is not a US-English keyboard, you must manually specify the keyboard type during installation. Otherwise, installation continues by using the default keyboard specification that is inconsistent with the system's actual keyboard type.
Workaround 1: If the system's keyboard is not a US-English keyboard, perform the following steps during installation:
When the Proposed Window System Configuration For Installation is displayed, press Esc.
The information on the Proposed Window System Configuration For Installation, which includes the keyboard type, is displayed only for 30 seconds. If you want to change configuration settings, you must press Esc before the 30 seconds lapse. Otherwise, the installation continues by using the displayed settings.
Change the keyboard type to the type that corresponds to your system's keyboard.
Press Enter to accept the changes and continue with the installation.
Workaround 2: If you want to change the keyboard type in a system that is already running Solaris 10 OS, use the kdmconfig program. Choose the option that applies to the type of X server your system is running.
If your system is running the Xsun server, follow these steps:
Run kdmconfig.
Use the Change Keyboard option to change the keyboard type.
Save the configuration.
If your system is running the default Xorg server, follow these steps:
Run kdmconfig.
Select the Xsun server.
Use the Change Keyboard option to change the keyboard type.
Save the configuration.
Run kdmconfig again to switch to the Xorg server.
The jfca driver for the following host bus adapters (HBAs) might cause system panics or I/O failures when these HBAs are connected to tape devices:
SG-PCI1FC-JF2
SG-PCI2FC-JF2
The jfca driver for these HBAs is prone to race conditions when certain operations are being run, and thus causes the errors. The operations are the following:
Link reset
Loop reset
Switch reset
Repeated link failures
Error messages similar to the following examples might be displayed:
I/O failure messages
jfca: [ID 277337 kern.info] jfca4: Sequencer-detected error. Recover immediately. last message repeated 18376 times jfca: [ID 716917 kern.notice] jfca4: ExgWarning: SendVerify(1): SHOULD ABORT THE ORIG I/O PKG=30007520bd8! scsi: [ID 107833 kern.warning] WARNING: /pci@1e,600000/SUNW,jfca@3,1/fp@0,0/st@w2100001086108 628,1 (st3): SCSI transport failed: reason 'timeout': giving up |
System panic message
panic[cpu1]/thread=2a100497cc0: BAD TRAP: type=31 rp=2a1004978d0 addr=a8 mmu_fsr=0 occurred in module "jfca" due to a NULL pointer dereference |
Workaround: Do not connect tape devices to either the SG-PCI1FC-JF2 or SG-PCI2FC-JF2 HBA.
A bus contention occurs if Quad Fast-Ethernet (QFE) cards share the same bus with any of the following adapters:
Sun GigaSwift adapter
Sun Dual Gigabit Ethernet and Dual SCSI/P adapter
Sun Quad Gigaswift Ethernet adapter
The infinite-burst parameter of the ce driver that is used by these adapters is enabled by default. Consequently, little or no bus time is available for the QFE ports that share the same bus.
Workaround: Do not place QFE cards on the same bus as the network adapters in the list.
The default timeout value for the SCSI portion of the SunSwiftTM PCI Ethernet/SCSI host adapter (X1032A) card does not meet the timeout requirements of Sun's SCSI DVD-ROM drive (X6168A). With marginal media, the DVD-ROM occasionally experiences timeout errors. The only exceptions are Sun Fire 6800, 4810, 4800, and 3800 systems. These systems overwrite the SCSI timeout value by means of OpenBoot PROM.
Workaround: For other platforms, use the on-board SCSI interfaces or DVD-ROM compatible SCSI adapters, such as the following examples:
X1018A (SBus: F501-2739-xx)
X6540A (PCI: F375-0005-xx)