HP Tru64 UNIX and TruCluster Server Version 5.1.B-4 Patch Summary and Release Notes (13156)

• Changes the way page migrations occur on a NUMA system to address poor performance
due to excessive paging.
• Corrects a problem with pagetable page allocations that can leave a thread waiting indefinitely
during a fork operation.
• Corrects the cause of "ubc_wire: hash failed" panics on non-NUMA systems.
• Correct the cause of "not wired" panics with System V shared memory and bigpages.
• Fixes an issue in the VM subsystem, wherein a page that is not managed by VM is incorrectly
identified as being managed by VM.
• Provides enhanced hwmgr show functionality for scsi path information.
• Fixes a problem in which crash dumps to Fibre Channel swap devices do not always succeed.
• Allows for control ports to be deleted by the hwmgr utility.
• Fixes a cluster deadlock/hang issue that occurs when a new device is discovered.
• Fixes a problem in which the “hwmgr refresh scsi -all” command does not always remove
stale paths.
• Corrects the cause of a process hang and system panic that occurs when using System V
shared memory and asynchronous I/O.
• Provides preferred path support for active-active, asymmetric storage devices.
• Improves tape read/write command reliability in SANs by allowing the use of FCP-2 Link
Level Error Recovery (LLER) for read and write tape commands if the destination Fibre
Channel port supports LLER.
• Enables Tru64 UNIX to work with tape devices that do not support non-tagged commands.
• Addresses a potential I/O performance bottleneck in which the tape driver may select the
same path for all tape drives in a system containing multiple HBAs and multiple tapes. With
this change, the tape driver can assign different paths to different tape drives to improve
tape I/O performance.
• Addresses a NUMA memory initialization issue that can lead to system instability.
• Allows systems with multiple paths to a large number of devices to boot faster.
• Fixes the cause of the "pmap_pagemove" panic that occurs with some third party drivers.
• Fixes a problem in which the changer code causes a kernel memory fault.
• Fixes an issue with the KZPCC backplane RAID adapter device driver (I2O) that causes its
logical disk drives to be identified as SCSI devices.
• Makes adapter error messages more descriptive.
• Corrects a memory leak.
• Fixes the problem with the hwmgr utility displaying the wrong output for network device
MTU.
• Addresses an issue in which the cam changer driver does not unwire user pages when a
path fails.
• Fixes a problem of system hangs that occur due to combined memory leak and DMA resource
leak in the EMX driver during the processing of an UNTAGGED request like device/lun
reset.
• Fixes a problem with KZPAC (SWXCR controller) that occurs when an open() issued on a
deleted unit succeeds and a read I/O request results in a system crash.
• Allows systems with multiple paths to a large number of devices to boot faster.
• Changes the way the aha_chim OSM driver handles timeouts. Previously, timeouts were
handled by registering a timer in the kernel callout queue for each I/O that timed out. As a
result, a burst of aha_chim timeouts could flood the kernel callout queue. The new method
times the I/O itself, so no entries are put in the kernel callout queue.
• Ensures that the alt firmware is stopped before freeing buffers when the alt card is
re-initializing.
• Fixes a problem with the alt driver regarding receiving 9000-byte packets.
3.4 Summary of Base Operating System Patches 69