HP Tru64 UNIX and TruCluster Server Version 5.1B-5 Patch Summary and Release Notes (March 2009)
• Fixes a Path MTU problem in network stack.
• Fixes double unlocking of netisr_lock in netisr thread.
• Resolves synchronization issues surrounding concurrent references to shared
memory.
• Fixes a SACK (Selective Acknowledgement) validation issue with TCP.
• Resolves a race condition between process exit and coredump.
• Fixes a problem in the FIFO code where usage of fuser(2) subsequent to opening
a FIFO may result in a kernel panic with the panic string VREF: Invalid
v_usecount transition.
• Fixes a problem wherein the signal mask was not getting restored after returning
from sigsuspend.
• Fixes potential hang or panic resulting from blocking under select_enqueue().
• Turns on the TCP Selective Acknowledgment (SACK) option by default.
• Corrects a problem with FIFO code where stat(2) or fstat(2) operation on
a FIFO returns NODEV in the st_dev field of the stat structure.
• Corrects an issue that can lead to a kernel memory fault panic when locking file
backed memory on a NUMA system.
• Corrects problems with RFC1323 TCP timestamps and PAWS implementation.
• Fixes a timing window in the VM subsystem and prevents a race condition between
the page fault and the pageout paths.
• Fixes an issue with NFS vnode reclaim path in CFS (Cluster File System)
environment.
• Fixes loss of multicast packet reception after netrain failover.
• Allows filtering of dup2 syscalls while audit object selection is enabled.
• Provides an enhancement to remove drawbacks of using TCP keepalives and make
LAN cluster more resilient.
• Corrects a potential security vulnerability whereby, under certain circumstances,
a local authenticated user could cause a Denial of Service.
• Fixes TCP SACK unaligned access panic.
• Improves performance of ufs extendfs operation on a mounted filesystem.
• Introduces a new sysconfig tunable to control the behavior of the O_APPEND flag
with respect to the pwrite(2) system call.
• Fixes a problem with the kernel traffic monitoring thread.
• Fixes a problem in FFM unmount code where, under certain circumstances two
or more unmount threads race and one thread can return without releasing the
NFS lock_for_exported results in a hang during unmount or shutdown.
• Corrects fixfdmn to display proper error messages when open()or write()
calls to the device fail.
• Fixes kernel memory fault panic in network stack.
• Corrects an issue with the TCP congestion control algorithm in Tru64 UNIX.
84 Tru64 UNIX Patches