Go to main content

Sun Fire X4470 Server Product Notes

Exit Print View

Updated: September 2018
 
 

Oracle Linux, SUSE Linux Enterprise Server (SLES), Red Hat Enterprise Linux (RHEL), and Oracle VM Known Issues

Table 10  Oracle Linux, SLES, RHEL, and Oracle VM Known Open Issues
Bug ID
Description
Oracle Linux 5.6 and RHEL 5.6 Fail to Allocate Memory Space on Onboard Intel 82576 Network Interface Card Ports
Issue:
Oracle Linux 5.6 and RHEL 5.6 display PCI: Failed to allocate mem resource for an onboard Intel 82576 network interface card, which supports the Intel SR-IOV feature. Oracle Linux 5.6 and RHEL 5.6 will then attempt to allocate memory space that is required by Virtual Function exported by Intel's SR-IOV capability.
Affected Operating Systems:
  • Oracle Linux 5.6

  • Red Hat Enterprise Linux (RHEL) 5.6

Workaround:
First enable SR-IOV support and ARI support features from the Advanced Menu in the BIOS Setup Utility.
15765632
Quad Gigabit Ethernet UTP PCIe and Dual Gigabit Ethernet SFP+ PCIe Option Cards Do Not Work on SLES 10 SP4 XEN
Issue:
On SLES 10 SP4 XEN, the Sun PCI-Express Quad Gigabit Ethernet Low Profile Adapter (X4446A-Z) and the Sun 10GbE SFP+ PCIe 2.0 Low Profile Adapter (X1109A-Z) do not work normally because of a XEN kernel issue.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 10 SP4

Workaround:
This issue has been fixed in SLES 11 SP1 XEN kernel; however, there is no workaround for this issue in SLES 10 SP4 XEN kernel.
15764667
Oracle VM 3.x.x and SLES 11 SP1 XEN Might Report map irq failed
Issue:
SLES 11 SP1 XEN and Oracle VM might report map irq failed.
Affected Operating Systems:
  • Oracle VM 3.x

  • SUSE Linux Enterprise Server (SLES) 11 SP1

Workaround:
Add the following to the kernel line of the GRUB file: extra_guest_irqs=64,2048 nr_irqs=2048
15790060
RHEL 6.2 - APEI: Cannot request iomem region <000000007f790800-000000007f790804> for GARs
Issue::
The following error message of APEI appears in RHEL 6.2 dmesg when the system boots using default kernel parameters:
APEI: Can not request iomem region <000000007f790800-000000007f790804> for GARs.
This inconsistent description of the APEI ERST feature in ACPI specification 4.x/5.x renders the ERST function of APEI unusable in Linux.
Affected Operating System:
  • Red Hat Enterprise Linux (RHEL) 6.2

Workaround:
Disable ERST of APEI by appending erst_disable to the kernel command line. The following information will be displayed in dmesg:
ERST: Error Record Serialization Table (ERST) support is disabled.
15546906
Storage Drive LEDs Are Incorrectly Illuminated When Using ICH10 On-board HBA With SATA Drives
Issue:
On the Sun Fire X4470 Server without a RAID PCIe card (with on-board ICH10 SATA controller), disk drive Fault and OK-to-Remove LEDs are incorrectly illuminated on hard disk drive slots.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
This issue has been addressed with an SLES 11 kernel update available from Novell. Refer to the Novell Documentation Administration Guide for information on obtaining updates using YaST Online Updater or use Novell patch finder at http://download.novell.com/patch/finder/
15584884
PCIe AER Driver Error Log Messages Are Incomplete
Issue:
The PCIe AER driver in SLES 10 SP3 and SLES 11 might produce excessive log messages that contain incomplete information about hardware errors.
Affected Operating Systems:
  • SUSE Linux Enterprise Server (SLES) 10 SP3

  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
For SLES 10 SP3:
Add the pci=noaer parameter as a boot argument to the kernel. To add the argument automatically at every boot, edit the GRUB boot loader configuration file /boot/grub/menu.lst. Add pci=noaer to the kernel line of the default boot title. For example:
title SUSE Linux Enterprise Server
root (hd0,2)
kernel /boot/vmlinuz pci=noaer root=/dev/disk/byid/ata-SEAGATE_ST95000NSSUN500G_0931M110L1_9SP110L1-part3 load_ramdisk=1 noresume showopts
initrd /boot/initrd
For SLES 11:
This issue has been addressed with an SLES 11 kernel update available from Novell. Refer to the Novell Documentation Administration Guide for information on obtaining updates using YaST Online Updater or use Novell patch finder at http://download.novell.com/patch/finder/
15606277
m1x4_core Driver for InfiniBand Fails to Load or Defaults to Single Pin Interrupt
Issue:
InfiniBand I/O devices using the mlx4_core driver and MSI-X interrupts might cause the driver to not load or to only load using IO-APIC interrupts.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
Install the SLES 11 kernel and OFED package update available from Novell. Refer to the Novell Documentation Administration Guide for information on obtaining updates using YaST Online Updater or use Novell patch finder at http://download.novell.com/patch/finder/
15608084
Error Messages Logged About Loading InfiniBand EHCA Driver
Issue:
If the OFED package group is installed, the system generates benign error messages about loading the InfiniBand EHCA driver. The errors are logged because the supported InfiniBand I/O device does not use the EHCA driver.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
Supported InfiniBand devices use the mlx4_core device driver. Ignore the error messages about the EHCA driver or disable attempts to automatically load the EHCA driver by editing the configuration file /etc/infiniband/openib.conf to change EHCA_LOAD=yes to EHCA_LOAD=no.
15608085
SLES 11 Produces Errors in Subnet Manager Start Script
Issue:
SLES 11 is unable to execute the subnet manager start script opensmd.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
This issue has been addressed with an SLES 11 kernel and OFED package update available from Novell. Refer to the Novell Documentation Administration Guide for information on obtaining updates using YaST Online Updater or use Novell patch finder at http://download.novell.com/patch/finder/
15608534
InfiniBand Driver Is Unable to Use MSI-X Interrupts
Issue:
The InfiniBand driver mlx4_core in SLES 10 SP3 cannot use MSI-X interrupts on systems with a high number of CPUs.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 10 SP3

Workaround:
Load the InfiniBand driver with MSI-X interrupts disabled, using one of the following two methods:
  • Issue the following command to load the driver with MSI-X interrupts disabled:

modprobe mlx4_core msi_x=0
  • Add the following line to the file /etc/modprobe.conf.local so that mlx4_core will never use MSI(X) when the driver is loaded:

options mlx4_core msi_x=0
15625335
SLES 11 Linux Kernel Panics When Attempting to Boot If No Memory or Nonworking Memory Is Installed in CPU 0
Issue:
The SLES 11 Linux kernel requires that Processor (CPU) 0 has memory installed and that the memory is in working order. If there is no memory for CPU 0, or if the memory for CPU 0 is unusable, the SLES Linux kernel panics at every attempt to boot.
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
For best results, update to the latest SLES 11 kernel, which will allow the operating system to boot with no memory in CPU 0 if you have memory installed elsewhere. Refer to the Novell Documentation Administration Guide for information on obtaining updates using YaST Online Updater or use Novell patch finder at http://download.novell.com/patch/finder/
Alternatively, you can repopulate the memory slots to ensure that CPU 0 has working memory.
15636521
Oracle Linux 5.x.x, RHEL 5.x.x, and SLES 11 SP1 Allow Only Limited Number of Interrupt Vectors
Issue:
Certain hardware configurations that include many high bandwidth I/O cards with drivers that use multiple MSI vectors per instance might cause the kernel to run out of interrupts. This situation might result in a long pause or system hang at boot time or I/O devices that are unusable due to driver instances unable to get required interrupts. This situation is especially likely to occur for the Xen Hypervisor and Dom0 with multiple InfiniBand and 10 GbE network cards using SR-IOV, but is also possible on the SMP kernel.
Affected Operating Systems:
  • Oracle Linux 5.1.0, 5.5, 5.6, 5.7, 5.8, 5.9

  • Red Hat Enterprise Linux (RHEL) 5.1.0, 5.5, 5.6, 5.7, 5.8, 5.9

  • SUSE Linux Enterprise Server (SLES) 11 SP1

Workaround:
Booting with kernel argument pci=nomsi will allow the system to boot with no long pause in UDEV and no system hang. The system will, however, be forced to use a single IO-APIC PIN interrupt per function. If you need MSI instead of IO-APIC interrupts, then you might need to use some combination of the following methods to alleviate the problem:
  • Blacklist some drivers by adding the module name to the text file /etc/modprobe.d/blacklist. This prevents the module from loading and using the device.

  • Force certain drivers (such as mlx4_core : infiniband) to load last, and take advantage of the code in those drivers that asks for optimal MSI number of vectors and, if those vectors are not available, retries with increasingly smaller number of vectors.

15643599
SLES 11 Does Not Handle MSIs Properly When VT-d Is Enabled in BIOS
Issue:
If VT-d is enabled in BIOS on a server running SLES 11, all interrupts go to a single processor (CPU).
Affected Operating System:
  • SUSE Linux Enterprise Server (SLES) 11

Workaround:
  • Use SLES 11, but in the /boot/grub/menu file, append nox2apic to the first line.