HP Tru64 UNIX and TruCluster Server Version 5.1B-6 Patch Summary and Release Notes
• Corrects AdvFS quota handling and enforcement to prevent EIO and false out-of-space
conditions and account for directory index blocks.
• Corrects the following problem with read-only and dual mounted file systems: If the system
crashes after a read-only dual mount, the next time the system is booted the AdvFS domain
panics on the files systems that were previously mounted as read-only dual mount.
• Fixes a kernel memory fault from imm_page_to_sub_xtnt_map().
• Fixes AdvFS AIO write error paths so the I/O completion steps are not repeated.
• Fixes a hang that occurs in file systems between racing memory-mapping threads.
• Turns a potential AdvFS panic into a domain panic.
• Corrects a problem with the AdvFS bitfile state where, under certain circumstances, the state
of the bitfile is changed without holding the lock that protects the state.
• Fixes an issue with vfast in which a vfast thread might cause a "kernel memory fault" panic
because of a race condition.
• Fixes a potential panic on an active domain doing heavy I/O while trying to create a new
fileset.
• Fixes problems that occurs when volume expansion (mount -u -o extend) races with other
code.
• Fixes a rare simple lock timeout during domain deactivation.
• Fixes a potential hang in AdvFS.
• Converts non-severe system panic calls to AdvFS domain panic calls while renaming files
under AdvFS.
• Corrects a problem in which an ENOSPC error is returned on a fragmented AdvFS file
system even though space is available. On a cluster, this can lead to a CFS WRITE ERROR.
• Fixes a premature out-of-space condition that can occur as a result of repetitively extending
the size of the volume.
• Fixes resource leaks seen after a device file gets revoked.
• Fixes a delete failure due to a “Disc quota exceeded (EDQUOT)” condition.
• Fixes a data inconsistency that could result from the failure of an I/O request at the CFS
Server due to exceeding the fileset quota.
• Improves AdvFS informational messages as follows:
— Advscan now reports if a domain has all of its volumes, but they are stored in a different
directories. This scenario will cause mount to fail.
— The AdvFS I/O error message now includes the location of a file that will help users to
translate the error number into an error message.
• Prevents race conditions that could cause a kernel memory fault while doing a migrate and
a rmvol on a striped file.
• Modifies the CFS flushing behavior during an rmvol.
• Helps reduce the size of extent maps of clone files cases where the original is modified
extensively under direct I/O.
• Enhances /sbin/AdvFS/tag2name command to print the name of the associated directory
when given the tag of an index file.
• Corrects a potential problem with modifying files via direct I/O when there is a clone fileset.
• Fixes a race during AdvFS volume removal that can cause a panic in the bs_osf_complete()
routine.
• Fixes a problem when monitoring I/O via AdvFSstat.
• Changes the behavior of migrate_normal and migrate_stripe when migrating an original
file that has a clone. If the clone was marked out of sync, migrate could come back with
E_CLONE_OUT_OF_SYNC even though the migrate succeeded. Now this case is caught,
and handled.
74 Tru64 UNIX Patches