ptx®/NFS® V4.6.4 is a subset of Sun Microsystems® NFSSRC V4.2, ported and enhanced to run on NUMA-Q® and Symmetry® systems. It supports filesystems larger than 2 GB and files up to 2 GB.
This version of ptx/NFS requires the following products:
For product compatibility with other DYNIX/ptx products, consult the DYNIX/ptx and Layered Products Software Installation Release Notes.
For information on installing ptx/NFS, refer to the DYNIX/ptx and Layered Products Software Installation Release Notes.
For ptx/NFS, you will need 3.2 MB in root and 8.8 MB in /usr.
For more information about ptx/NFS and NIS+ products, refer to the following manuals:
This section lists fixed and open software defects that have been reported against this release of ptx/NFS.
The following software defects are fixed in ptx/NFS V4.6.4:
(239868, 240053) Lock requests were dropped by an application that did a lot of file locking.
(240829) A lock-contention bottleneck occurred in nfs_kmem_free.
(242140) mount_nfs returned a 0 exit status when it failed, which resulted in confusion when shell scripts attempted to verify if the mount_nfs was successful.
(248238) mount_nfs did not check the length of server and path names. This was liable to result in a user providing a path or server name that was too long, causing a buffer overflow.
(248275) The -o force option was added to the mount_nfs command to pass the correct parameters to the kernel for a forced mount.
(248292) A bad size option in getkerndata caused rpc.rstatd to fail with the message:
nnnnnnnn time tolog/nocons/console - getkerndata
nnnnnnnn time tolog/nocons/console - Not enough space
nnnnnnnn time tolog/nocons/console - Nengine return value is 12
(248820) A forced NFS unmount operation left the old mount structures in memory. Additional functionality has been added to prevent this from occurring.
This section describes known software defects in this release of ptx/NFS. The number that appears in parentheses in the title of each problem report is the problem-tracking-system number assigned to the report.
The ptx/ADMIN menus used to manage the automounter maps create file names with periods instead of underscores, creating /etc/auto.master instead of /etc/auto_master.
Workaround. None.
ptx/ADMIN returns an error when a user tries to add more than 10 hosts in the "Enter list of hosts to export this filesystem read/write to" section of the "Exported filesystems" menu. The error is:
Too many members in the read/write group
Must be less than 10. Operation aborted.
Workaround. If you wish to place more than 10 hosts in the read/write group, edit the /etc/exports file manually and then run exportfs.
The automounter is not able to read maps from NIS servers.
Workaround. None.
rpcinfo is unable to communicate with previous versions of the portmapper.
Workaround. None.
The mount_nfs command returns a 0 exit status when it fails because the mount point is already in use. This can cause problems when mount_nfs is used by shell scripts that try to verify whether the command was successful.
Workaround. None.
ptx/NFS limits itself to one-half the total buffer cache. This does not work correctly on multiquad systems, where one quad can run out of buffers.
Workaround. None.
The mount_nfs command does not validate the options that are passed to it and passes bad options to /etc/mnttab.
Workaround. None.
Environment macro expansion in the automounter does not work correctly.
Workaround. None.
One of the options in the automounter is to start up the automounter server with a known port number and replace a crashed automounter with the same port number. However, the automounter refuses to start up with the -p option.
Workaround. None.
When you create an automount setup through ptx/ADMIN and use direct or indirect maps, the option for the type of mount is positionally dependent. If the mount type is last on the line, it will generate errors in the ktlog.
Workaround. None.
The -f, -p, and -l options to the automount command are not documented in the man page.
Workaround. None.
Comments in the auto.master file are not ignored by the automounter, which results in errors in the ktlog.
Workaround. None.