VMware ESX Server 3.5, Patch ESXe350-200802401-I-BG: Firmware Update

Summaries and Symptoms

This patch provides fixes for the following issues:

• Fixes for link state and packet latency issues seen on certain ESX Server hosts.
  
  On some systems such as IBM BladeCenter using copper pass-thru modules (CPM) or HP C-class 7000 Blade, the tg3 driver misreports the link state of its port. On IBM BladeCenter using CPM, when the patch cord is disconnected from the physical switch, ESX Server logs the downed link properly, but ping latency increases from 1 millisecond to 100-200 milliseconds.
  
  
• Enhanced validation checks prevent potential memory corruption problems.
  
  No customer-visible symptoms.
  
  
• Minor edits to Storage VMotion error messages.
  
  All occurrences of "DMotion" in error messages have been changed to the correct term, "Storage VMotion."
  
  
• Resolves DMA issues on LSI 1078 and 106XE HBAs.
  
  A hardware issue in LSI 1078 chip-based controllers results in the improper mapping of the DMA address, which results in data corruption. LSI 106XE controllers (SAS1064E and SAS1068E, B1 and earlier revisions) are not able to perform DMA for addresses beyond 4GB.
  
  
• Due to a possible race in the device file system code, ESX Server might crash when loading and unloading storage drivers.
  
  ESX Server crashes when it loads and unloads storage device drivers. Incorrect output might appear when you run vdf –h /vmfs/devices to determine the current disk space.
  
  
• Fixed an issue in the skb_copy implementation in vmklinux, which is a module that allows Linux device drivers to run in the VMkernel.
  
  The skb_copy() function of the vmklinux module does not work as expected. In rare cases, the transmit packets are corrupted.
  
  
• The REPORT LUNS SCSI command from a guest operating system sometimes returned an incorrect value.
  
  A malformed REPORT LUNS command (for example, possibly created by a faulty application or SCSI utility in the guest operating system) might sometimes return incorrect status.
  
  
• ESX Server printed out spurious warning messages in /var/log/vmkwarning on systems with mixed stepping CPUs, which include those with dual-core Intel Xeon 5100-series (Woodcrest) and quad-core Intel Xeon 5300-series (Clovertown) processors.
  
  On systems with mixed stepping CPUs, /var/log/vmkwarning displays messages that look like the following:
  
  vmkernel: 0:00:00:00.XXX cpuX:XXX)WARNING: Cpu: SaveAndCheckCPUID:518: version 000006fb but BSP 000006f7
  
  These messages are harmless and can be ignored. This patch version moves the messages from warnings in /var/log/vmkwarning to informational messages in /var/log/vmkernel.
  
  
• Simultaneous rescans and RDM creations, RDM opens, and RDM deletions can sometimes fail.
  
  While performing simultaneous rescans and hot addition or deletion of RDMs (or other such operations that involve RDM creation, RDM open, or RDM deletion) the procedure can sometimes fail. The system reports the exceptions "concurrent modification" and "file not found" in /var/vmware/log/hostd.log. After applying the patch, the system might still report the "concurrent modification" exception. If this occurs, retry the operation. You should not see the "file not found" exception after applying the patch.
  
  
• The file packages.xml marks the update as done, even though it is aborted.
  
  The firmware version returned after an incomplete update is incorrect.
  
  
• The VirtualCenter agent (vpxa) sometimes crashes during the discovery of a LUN.
  
  When using VirtualCenter to perform an ESX Server host rescan to discover new LUNs, the VirtualCenter agent (vpxa) crashes and the following message is logged because the secondary host cannot be reached:
  
  An error occurred while communicating with the remote host.
  
  Note that the rescan operation is successful in this case and that the VirtualCenter agent (vpxa) is automatically restarted.
  
  
• Various requests for system information (such as system health status, BIOS info, system identification, and hardware status) invoked i18n code and crashed hostd.
  
  The HA cluster becomes unresponsive. The VirtualCenter agent (vpxa) is running, but its daemon (vpxd) errors out when it tries to communicate with vpxa because the communication goes through the vmware-hostd proxy. Connecting to hostd with the VirtualCenter client also fails. Killing hostd fixes the problem.
  
  
• Virtual machines (typically Windows) under heavy disk workload crashed with a blue screen, because of on-disk corruption during the consolidation of snapshots.
  
  While running performance and benchmarking applications like IOzone or MemHog, and while performing snapshot consolidate operations, virtual machines crash with a blue screen that reports KERNEL_STACK_INPAGE_ERROR or SYSTEM_SERVICE_EXCEPTION (77).
  
  
• Fixes an issue which could potentially crash hostd while adding or reconfiguring a virtual machine SCSI device.
  
  Adding or reconfiguring a virtual machine SCSI device might crash hostd.
  
  
• ESX Server 3.5 hosts using more than two processor sockets cannot be licensed when using host-based licensing.
  
  Licensing an ESX Server 3.5 host with more than two processors fails when attempting to use host-based licensing regardless of how many licenses are available. The ESX Server software installs properly, but the server fails to leave the unlicensed mode and instead runs in evaluation mode.
  
  
• Fixed a flaw in how the aacraid SCSI driver checked IOCTL command permissions.
  
  This flaw might allow a local user on the service console to cause a denial of service or gain privileges. Thanks to Adaptec for reporting this issue. The Common Vulnerabilities and Exposures project (cve.mitre.org) has assigned the name CVE-2007-4308 to this issue.
  
  
• Updated the support script (vmware-support) so it collects the oem.tgz file. Collecting this file helps VMware recognize and diagnose reported issues.
  
  No customer-visible symptoms.
  
  
• On certain Dell hardware configurations, SNMP traps show incorrect drive numbers due to incorrect generation of object paths.
  
  SNMP traps show the incorrect drive number.
  
  
• On certain Dell hardware configurations, CIM reports too many drive sensors.
  
  Drives on external storage are enumerated in reverse order in VI Client.
  
  
• Provide a correct content ID (CID) value for thin-provisioned virtual machine disks after they are moved using Storage VMotion.
  
  Storage VMotion fails with the error message: Failed to unstun VM after disk reparent. The virtual machine is partially migrated and powered off. Generally, the virtual machine cannot be powered on again. For more information, see KB 1003874 at http://kb.vmware.com/kb/1003874.
  
  
• Fixes ESX Server support for Intel Zoar NICs on a VLAN trunking configuration. A flaw exists between the ESX Server networking subsystem and the igb driver.
  
  When implementing a VLAN trunking configuration on an ESX Server system that utilizes the igb driver for Intel Zoar NICs, connectivity is lost.
  
  
• Host hangs after Emulex timeout errors appear when Emulex HBA cannot transfer more than 1MB by means of direct memory access (DMA) in an AMD chipset environment.
  
  While running Iometer on a Windows virtual machine whose storage is on a SAN-based LUN, the ESX Server host times out, loses its connection to LUNs, and the system stops responding. A reboot fixes the issue temporarily. The ESX Server host is a Sunfire X4200 using a Sun-branded Emulex HBA.

Deployment Considerations

None

Patch Download and Installation

For information on using VMware Update Manager to automatically update ESX Server 3i hosts, see the VMware Update Manager Administration Guide.

ESX Server 3i hosts can also be updated by downloading the most recent patch bundle from http://www.vmware.com/download/vi/vi3_patches_3i.html and installing the bundle using the Remote Command Line Interface (RCLI). For more information on using the RCLI and the vihostupdate command to update ESX Server 3i hosts, see the ESX Server 3i Configuration Guide and the ESX Server 3i Embedded Setup Guide or the ESX Server 3i Installable Setup Guide.

Note: ESX Server 3i hosts do not reboot automatically when using the RCLI and the vihostupdate command. The hosts must be manually rebooted after any update has been installed. Please refer to the guides listed above for more information or for information on how to perform updates so that they can be rolled back after installation.