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

• Resolves several DECThreads faults and resolves performance issues with certain Java
applications.
• fixes memory leak problem in cron.
• Raises the highest ID handled by edquota and repquota and converts the repquota program
to using the GETQUOTA/GETQUOTA64 quotactl's for obtaining data, instead of directly
reading from the quota file. This results in more recent data and, in a cluster, consistent data
across all nodes. It also enables handling the highest UIDs (up to 4294967294).
• Fixes an awk() argument buffer limitation that can cause patch installations to fail if a large
number of subsets (in excess of 500) are loaded on the system.
• Fixes sh to handle a large number of open files.
Patch 27010.00
OSFBIN540
• Makes possible a sticky connection feature for cluster alias.
• Updates sysconfig to use the cluster interconnect to allow for greater SSI collaboration, which
will help with changing variables on hung systems, single user systems, and normal running
systems.
• Allows the UFS attribute delay_wbuffers to be tuned using sysconfig.
• Allows the packet filter variables pfilt_loopback and pfilt_physaddr to be tuned using
sysconfig.
• Increases the default value of ipqmaxlen (IP input queue) to 2048.
• Provides an option to enable cluster NFS clients to use a nonprivileged TCP port to check
to see if a remote NFS server is up.
• Corrects a potential issue with NFS version 3 memory mapped files that can lead to a system
panic.
• Allows the stat system call to correctly report the st_blocks for dvdrom files.
• Corrects a condition that causes a panic while creating or extending large UFS file system.
• Corrects a problem in which under certain load conditions shared memory usage can lead
to an inconsistency that results in a "u_ssm_oop_deallocate: reference count mismatch"
panic.
• Corrects a potential hang on exit from applications utilizing /dev/poll.
• Fixes a condition in the kernel whereby an incorrect internal status can be returned from
mpsleep(), thereby causing potentially incorrect behavior.
• Fixes problems in tcp_output that cause connections to hang when window fills and fixes
a potential loss of data when a connection is closed.
• Corrects a potential loss of data a connection is closed.
• Corrects inappropriate TCP probe timeouts associated with case.
• Adds support for CPU offline on GS1280 systems (required for Capacity on Demand).
• Corrects a problem with the keepalive mechanism that causes TCP connections to disconnect
unexpectedly.
• Corrects a problem in which changing memory protection results in a kernel memory fault
panic.
• Fixes a condition that can cause a panic in the kernel during interconnect operations involving
configuration requests.
• Corrects a problem in which an invalid core file may be generated following abnormal
program termination.
• Fixes mount/umount failures and panics in FMS, UFS FDFS.
• Fixes a situation in which mmap memory locked with mlockall() using the MCL_FUTURE
flag does not become wired automatically.
3.4 Summary of Base Operating System Patches 91