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

• Fixes an AdvFS path that can cause a panic in the advfs_page_busy() routine.
• Fixes a deadlocking problem in the kernel where a file open on a clone could hang when
ACLs are enabled.
• Fixes a hang that can occur during the renaming of an AdvFS file.
• Displays the correct error message for freezefs -q on a non-AdvFS file system.
• Adds comment to reserve 0x10000000 and 0x20000000 for AutoFS flags.
• Prevents a O_DSYNC write failure under the following situation:
1. The user creates a new file.
2. Closes the file.
3. The vnode for the file is recycled.
4. Reopens the file with the O_DSYNC flag.
5. Writes to the file, overwriting already allocated storage.
6. The write from step 5 returns to the application.
7. The system crashes.
• Corrects the NFS server's handling of files open for direct I/O.
• Fixes an internal problem in the kernel's AdvFS, UFS, and NFS file systems where extended
attributes with extremely long names (greater than 247 characters) could not be set on files.
The new limit is 254 + a null string terminator.
• Corrects a problem that can produce the following symptoms:
— When producing a clone of a file (that is, a fileset has been cloned and one of its files is
being written to) or when migrating a file (defragment, migrate, balance, rmvol),
directories lose attributes and are seen as files, and files lose their ACLs and other
attributes under the following circumstances:
◦ On multi-volume domains with a volume that is out of space
◦ By the use of a property list on a multi-volume domain
— The rmvol command enters into an infinite loop while trying to move a file from one
volume to another.
• Fixes a problem where threads can hang when performing a malloc() function.
• Prevents a kernel memory fault panic in _OtsMove when going through the fs_read() routine.
• Prevents a potential hang during a umount if a domain_panic has been encountered.
• Provides a workaround for a domain panic when a data inconsistency in the deferred-delete
list of an AdvFS file system is detected.
• Corrects idle-wait time accounting within the AdvFS file system, as reported by vmstat -w.
• Corrects a condition in which command response deteriorates to about 30 seconds when
doing an rmvol on a domain with a volume containing large files (8 GB).
• Removes the obsolete function bs_bfdmn_flush_all.
• Prevents a potential unaligned memory crash when ACLs are on.
• Eliminates the lock_terminate: lock owned panic.
• Fixes a problem where data from an AdvFS file with a frag could be written to an incorrect
location if an NFS client grew the file.
• Fixes an AdvFS asynchronous direct I/O problem that can cause a thread to hang.
• Fixes a problem encountered where a truncated AdvFS file erroneously zeros data for the
remaining leading segment of the file.
• Corrects a condition that causes a panic resulting from a kernel memory fault in
access_invalidate.
• Corrects a problem in mount or domain activation after a panic, where a fileset (domain)
cannot be mounted without running fixfdmn.
• Improves performance for CFS filesets mounted with the server_only option. A log sync for
create transactions is not needed for such filesets.
3.4 Summary of Base Operating System Patches 79