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

• Provides a tunable attribute that gives system administrators the ability to adjust the weight
that the kernel places on the NUMA locality for forks on AlphaServer GS80, GS160, GS320,
and GS1280 systems.
• Sets a software limit to prevent serious performance problems that can occur when TCP
connections that have an rate-limit enforced by a downstream network device overrun the
device.
• Adds per-binary big page controls to complement the system-wide tunable attributes.
• Enhances HP-XP array controller support and possible future new tape device support.
• Fixes a panic condition in AIO.
• Fixes a "simple lock: time limit exceeded" panic.
• Fixes a problem that prevents access to the AutoFS file systems if ACLs are enabled.
• Improves NFS client performance on NUMA systems by doing a better job distributing data,
locks, and nfsiod threads across Resource Affinity Domains (RADs).
• Addresses the dispatching of NFS server requests for the public file handle on cluster
members and multi-RAD systems.
• Fixes how the NFS client handles full NFS version 3 64-bit file IDs.
• Resolves a problem that prevents the viewing of files created by third-party software on
some CD-ROM media.
• Resolves a problem of not being able to view files on some CD-ROM media that is created
by third-party software and corrects the erroneous reporting of success when attempting
to write beyond the file size limit using synchronized I/O and the calculation of
_PC_FILESIZEBITS, which is used by the operating system for pathconf file characteristics.
• Fixes a race condition in the kernel AIO code that can panic the system with either a kernel
memory fault or a duplicate malloc free.
• Corrects a potential panic with large memory processes using System V shared Memory at
process exit.
• Corrects a panic resulting from a race condition between vnode deallocation logic and the
use of CACHE_LOOKUP_REF/CACHE_LOOKUP_RELE in grab_bsacc and bs_dealloc_access
(which attempt to block vnode deallocation).
• Fixes a memory leak in the NFS server encountered when it receives malformed packets.
• Fixes an AutoFS panic during an unmount operation when AutoFS tries to remove a directory.
• Corrects a problem in which Tru64 UNIX sees an HP-XP RAID array controller as a disk
after an HP-XP storage device is added to the system.
• Corrects a problem in which table() calls were not correctly getting process arguments.
• Improves the scaling of IP reassembly code on large SMP machines. NFS servers are especially
susceptible when a large number of clients attempt to write at the same time.
• Implements buffer cache page checksum caching for NFS client pages.
• Fixes a problem that can cause a system crash when an NFS server exports files on a
third-party file system (that is, one not built into Tru64 UNIX).
• Fixes a rare kernel panic that occurs during the handling of a clock tick when class scheduling
is used.
• Corrects a potential process hang that occurs while exiting a system that has dynamically
powered off a processor.
• Corrects a system panic when running with big pages enabled.
• Addresses problems that occur when taking noninteractive core dumps using the coredump
command.
• Corrects potential hangs of applications using the pshared subsystem that can occur as a
result of a thread failing to wake up after its condition variable timer had expired.
• Fixes two potential problems in the NFS V3 client in which unstable writes can remain
uncommitted when they should have been committed to stable storage.
3.4 Summary of Base Operating System Patches 97