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

• Corrects the cause of a panic on a cluster system when renaming a root specified by the
chroot command.
• Fixes a kernel memory leak that occurs during routing table updates, which over time, can
cause memory shortages for machines attached to a constantly changing network, as in the
case of mobile IP.
• Fixes the cause of a hang/panic that occurs soon after a console printf returns the message
“NFS over TCP client xxx not reading replies, continuing."
• Corrects a kernel memory fault in exit() where uu_curinfo is nil.
• Fixes MIP6_CACHE_READ_TO_WRITE_LOCK() to not call lock_read_to_write() when in
lockmode=0 to avoid a panic.
• Modifies asynchronous I/O to prevent it from releasing file region locks.
• Fixes "issig recursion" panics.
• Fixes a simple lock timeout panic in the ubc_invalidate routine.
• Corrects a condition in which the fork() function can lead to a “vm_pg_free: page mapped"
panic when big pages are enabled.
• Corrects a problem with HSZ events that cause I/O errors rather than doing the appropriate
retries so the I/O can be successful.
• Corrects problem that prevents proper XP RaidManager operation.
• Corrects a Routing Header Type 2 length miscalculation.
• Fixes a problem in which some reads on /dev/random wait until the entropy pool is
completely full while other reads of /dev/random and /dev/urandom do not.
• Fixes a problem on a VMAC -enabled cluster wherein the member that acts as the
Proxy-ARP-Master for the default cluster alias address sometimes incorrectly uses the VMAC
address created for it as the source MAC address.
• Fixes the cause of a "u_anon_unlock_page: anon already unlocked" panic.
• Fixes a problem in which the NFS version 3 server can reset the setuid and setgid bits of a
file or directory after attempting to comply with the specified attributes in the request.
• Addresses a performance issue seen in certain applications that allocate and exercise a
memory segment close to the system's bcache size.
• Prevents aio-related slowdowns.
• Corrects an internal error so that UBC invalidate operations can execute correctly.
• Fixes the cause of an “mcs_unlock: lock not currently owned” panic.
• Addresses problems with reading third-party file systems exported by the NFS server.
• Fixes crashes seen on clustered systems.
• Addresses the panic “pmap_enter_bigpage: attempt to map big page in kernel map” that is
sometimes seen on large AlphaServer GS1280 configurations.
• Fixes an AutoFS problem,where indirect-mapped key directories are not removed for failed
mounts.
• Fixes a potential deadlock hang between a truncate system call and a read system call on a
clustered system.
• Enables AutoFS mount-on paths to have the correct maximum length when AutoFS files
are mapped directly.
• Fixes a condition that causes the evmwatch or evmshow commands to display incorrect
IPv6 addresses.
• Corrects a problem with handling of null symbolic links.
• Exports the creation time of a process via its /proc entry's ctime attribute.
• Fixes a problem in which questionable behavior by non-TRU64 NFS clients results in NFS
over TCP server threads hanging.
• Fixes a problem related to the deletion of binding cache entries in mobile IPv6, which affects
multi-RAD systems like the AlphaServer GS1280.
3.4 Summary of Base Operating System Patches 99