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

• Fixes a panic condition in the NFS client code resulting from the mishandling of unaligned
data on clustered systems.
• Corrects a kernel memory fault that results from the de-referencing of a null processor
pointer encountered when auditing a network event after having sent a SIGKILL to the
auditd daemon.
• Forces the use of a non-cluster interconnect address in audit records.
• Fixes a kernel memory fault panic in IPv6, and corrects improper or leaked reference counts
with IPv6 route entries.
• Fixes a panic that occurs when using the TCP SACK option.
• Fixes a kernel memory fault panic that occurs when the vm tunable anon_rss_enforce is set.
• Fixes a socket kernel memory leak.
• Fixes a race condition in the UBC.
• Corrects tracking of controlling terminal reference to session structure.
• Corrects a problem that causes the transmission of duplicate FIN packets, which could result
in a stuck connection.
• Allows systems with multiple paths to a large number of devices to boot faster.
• Fixes an issue with the NFS client async daemon that can occur in specific NUMA
configurations. These configurations would have RAD numbers that are not part of the
partition but within the range of valid RAD numbers for that partition.
• Fixes an fsync (NFS/UFS) operation that failed to always flush all dirty pages.
• Fixes a “lock_done: lock not currently owned” panic that occurs early in a boot.
• Resolves lock management issues within the UBC that can lead to “mcs_unlock: current
lock not found” and “mcs_lock: time limit exceeded” panics.
• Removes NFS warning message about malaligned RPC messages.
• Makes the poll() function compliant with revised UNIX98 standards.
• Fixes a hang caused by the Async I/O subsystem.
• Corrects a problem with the mount command in which issuing a mount -l command lists
the property of a dual mounted fileset incorrectly as nodual.
• Fixes a problem that causes some TCP connections to reset when sending large amounts of
data (more than 2GB) to a very efficient receiving host.
• Corrects the incompatibility of the waitpid() system call with revised UNIX98 standards.
• Fixes an internal kernel declaration for variable maxuthreads that resolves the following
types of kernel failures when the sysconfig value for max_threads_per_user in the proc
subsystem is set to a value larger than 2147483647:
thread_create() failed for pid # : maxuthreads (=-1) exceeded for uid 201
thread_create() failed for pid 1479950: maxuthreads (=-#) exceeded for uid 15
• Fixes a panic resulting from CDFS using a deallocated vnode.
• Corrects a problem in which the tmt_walk_list() function in nifftmt.c mallocs memory for
the tmt_state_info array, but fails to free memory when thread terminates.
• Reduces the default value of IPFRAGTTL from 60 to 18 to avoid reassembly problems.
• Modifies the setluid system call to handle the license correctly when a non-root user switches
to different non-root user using login command.
• Corrects a problem in which elevated load averages were reported on NUMA class systems.
• Fixes resource leaks seen after a device file is revoked.
Patch 27011.00
OSFBINCOM540
66 Tru64 UNIX Patches