HP Tru64 UNIX and TruCluster Server Version 5.1B-6 Patch Summary and Release Notes

A mismatch between the value reported by the write system call and the number of
bytes written.
—
— Unavailable and unused storage.
• Causes the deallocation of preallocated storage that the caller is not using.
• Improves the flushing of the AdvFS log.
• Fixes an error that can cause a multivolume domain to report ENO_MORE_BLKS when
some volumes still have free storage.
• Fixes a condition that causes a kernel memory fault.
• Fixes a condition that causes a system hang that occurs when the rename command is called
with "." as the target. This patch also reinforces other rename argument restrictions.
• Fixes an RBMT metadata inconsistency that prevents a file system from being mounted.
• Changes a system panic resulting from a kernel memory fault in imm_remove_page_map
into a domain panic.
• Fixes a condition that can cause the invalidation of mmap dirty pages before they can be
flushed to disk.
• Improves the performance of systems performing heavy file I/O.
• Corrects a potential deadlock hang between a truncate system call and a read system call
on a clustered system.
• Fixes an AdvFS alignment fault panic caused by inconsistent AdvFS metadata in a directory.
In particular, the directory's entry size is an unaligned value.
• Fixes erroneous logic to ensure that a domain panic on the cluster_root results in a regular
panic for the cluster node on which the domain panic occurs.
• Corrects a condition that causes an "mcs_lock: no queue entries available" domain panic.
• Corrects a problem in which the cp -p command will not copy DMAPI-managed region
information.
• Helps to avoid a kernel panic (kernel memory fault) during a vfexer test.
• Fixes the cause of a system crash when running with lockmode=4. This correction avoids
the following types of panics that result from quotactl requests on AdvFS filesets:
mf in dyn_hash_remove
lock_terminate: lock held
• Eliminates the “neg refCnt panic” in AdvFS.
• Fixes a potential panic when using the rmfset command.
• Fixes an infinite loop hang that occurs under special circumstances.
• Fixes an E_TOO_MANY_ACCESSORS error that can occur when deleting a clone fileset.
• Fixes a kernel memory leak that occurs when vfast is in use.
• Fixes a cluster hang where one node tries to get a DIO token and another node tries to start
a transaction while a third thread is waiting for the clu_clonextnt_lk lock.
• Fixes a potential deadlock hang between a truncate system call and a read system call on a
clustered system.
• Prevents a vfast thread from using too much CPU when scanning the AdvFS sbm.
• Provides corresponding memory frees to various mallocs in AdvFS.
• Corrects an infinite looping condition in a vfast thread.
• Corrects a problem in which I/O error codes were not always propagated correctly when
AdvFS directIO was used without AIO.
• Fixes a situation that occurs on a full file system in which a write using directIO via AIO
may report the incorrect number of bytes written.
• Fixes an unaligned access panic in insert_seq().
• Fixes a check for an invalid lookup operation through the . tags interface and prevents an
AdvFS domain panic.
Summary of Base Operating System Patches 77