8    New and Changed Features in Version 5.0

This chapter provides brief descriptions of features that were new to the Tru64 UNIX operating system in Version 5.0 or that changed significantly from previous releases.

Tru64 UNIX Version 5.0 is a major release that includes the following enhancements:

The following sections provide more information on the major new features included in Version 5.0.

8.1    Scalability Enhancements

Version 5.0 provides the following scalability enhancements:

In addition to these enhancements, this version of Tru64 UNIX also provides extended system limits. More information on some of these limits can be found in the relevant sections of this manual. For more information on system limits in general, see the Software Product Description (SPD).

8.2    Installation Improvements

This release provides several improvements to the Update and Full Installation processes.

The following enhancements have been made the Update Installation process:

The following improvements have been made to the Full Installation process:

For more information, see the Installation Guide.

8.3    System Management

This version of Tru64 UNIX provides many new features and enhancements to its system management utilities. The following sections highlight some of these changes. For more information, see the System Administration guide, the Installation Guide, the sysman_intro(8) reference page, and the online help for the applications.

8.3.1    Event Manager

The Event Manager (EVM) has been added to the SysMan utilities. EVM provides a centralized means of gathering, distributing, storing, and reviewing event information, regardless of how the events are posted. Event Manager makes event information more accessible and provides an event infrastructure that is flexible and adaptable. Some of the features included with EVM are:

For more information, see the System Administration guide and the EVM(5)reference page.

8.3.2    Exporting Tru64 UNIX System Management Data Using SNMP

This release includes methods for exporting Tru64 UNIX system management data using the Simple Network Management Protocol (SNMP). For more information, see the System Administration guide.

8.3.3    SysMan Integration and Application Access

The integration of and access to the SysMan applications has been enhanced in this version of Tru64 UNIX. You can now access the SysMan applications from the System_Admin section of the desktop's Application Manager, the Custom Setup checklist, and by using the sysman command.

The sysman command provides access to the suite of system management utilities, making it easier for you to find the applications you need to manage your system. This suite includes the SysMan Menu, the SysMan Station, the sysman command line interface, and the sysman -clone command. The following list provides a brief overview of these applications:

For more information on system management applications, see the sysman(8) reference page, the online help for the specific application, and the System Administration guide.

8.3.4    New and Changed SysMan Configuration Applications

This version of Tru64 UNIX provides several new configuration applications and many enhancements to existing system management applications. The following list provides a brief overview of the new applications and the changes to the existing ones. For more information on these applications, see the online help for the application and the Network Administration and System Administration guides.

8.4    File Systems and Storage

The enhancements to file systems and storage include the following:

8.4.1    File Systems Support for Increased Subdirectories and Links

This version introduces UFS Version 4, which simply supports 65533 hardlinks and continues to support existing UFS (now called Version 3). There is no on-disk structure change; however, there is an incompatibility problem when you use UFS Version 4 with earlier versions of the operating system. For more information, see the System Administration guide.

AdvFS, UFS, and VFS now support up to 65531 subdirectories. Previously they supported 32765 subdirectories. The maximum number of links to a file has been increased from 32767 to 65533.

AdvFS domains created under previous releases are called AdvFS Version 3 domains, while AdvFS domains created under Tru64 UNIX Version 5.0A are AdvFS Version 4 domains. An AdvFS Version 4 domain allows up to 65531 subdirectories and up to 65535 links to a file. All AdvFS Version 3 domains, even those created using the Tru64 UNIX Version 5.0, or higher, mkfdmn -V3 command, will support, at most, 32765 subdirectories and 32767 links to a file.

Because AdvFS Version 3 domains cannot have more than 32767 links, they can be moved back and forth between Tru64 UNIX Version 5.0, or higher, and earlier versions of the operating system. AdvFS Version 4 domains, however, will only work on Tru64 UNIX Version 5.0, or higher, systems.

Note that the pathconf() and fpathconf() system calls will return the maximum value for a particular pathname with the _PC_LINK_MAX option.

For more information on AdvFS, see the AdvFS Administration. For more information on UFS, see the System Administration, and the newfs(8) and fsck(8) reference pages.

8.5    New Networking Features

Networking enhancements include the following:

8.6    DAPBA and DAPCA ATM Adapter Support

This release of Tru64 UNIX contains support for the following three new ATM adapters:

These adapters are all 64-bit PCI cards. See the lfa(7) reference page for more information on the adapter and the sys_attrs_lfa(5) reference page for more information on the driver configuration options.

Not all platforms support these adapters. See the Server Options Catalog for the number and type of each adapter supported on your platform and for any configuration rules that apply.

8.7    Advanced Server for UNIX

The integration of the Advanced Server for UNIX with the Tru64 UNIX operating system has been enhanced to provide for Single Task Management capabilities for UNIX and Windows/NT systems. The UNIX administrator can use UNIX or Windows NT administration tools to configure resources for PC users, such as the following:

8.8    Enhanced TruCluster Support

Several features included in this version of Tru64 UNIX provide support for future TruClusters products, in addition to providing other functionality. These features include the following:

For more information on new device names and CDSLs, see the System Administration guide.

8.9    Documentation

The Tru64 UNIX documentation set has been updated to make it more comprehensive and easier to use. Some of the major improvements are:

8.10    Additional Changes

The following sections provide brief descriptions of additional changes included in Tru64 UNIX Version 5.0.

8.10.1    Changes to Device Naming

This release provides full support for FibreChannel, SCSI-3, and wide SCSI devices. The implementation of this support required a major change to the Tru64 UNIX device naming scheme. This version of Tru64 UNIX implements this change for all disk and tape devices.

For example, prior to Version 5.0, disks were named as follow:

This naming had encoded within it the bus and Logical Unit Number (LUN) of the SCSI disk. For example, disk 0 on bus 0 was rz0. Disk 0 in bus 1 was RZ8, and so on. As a result Tru64 UNIX was limited to supporting no more than eight devices per bus, because the name for any additional devices would collide with other devices.

Wide SCSI supports up to 16 devices per bus; within FibreChannel the number is in the thousands. FibreChannel also allows the LUNs to change dynamically, which the old device naming scheme could not support.

Therefore, in Version 5.0 disk names have the following format:

Tape drives have the following format:

(For more information on the formats, see the System Administration guide.)

The new device name will use the world-wide identifier (WWID) of the disk. A disk's WWID is set by the manufacturer for devices that support it and is unique. Therefore, no two disks can have the same WWID.

Using the WWID to identify a disk has two implications:

Tapes devices will reside under the /dev/tape directory; no-rewind tape devices will reside under the /dev/ntape directory. This version of Tru64 UNIX supports the existing device names as a compatibility option, but the same device cannot be accessed through both the old and new name at the same time.

The following utilities have been added to enhance the support for device naming and hardware management:

8.10.2    Changes to Data Structures to Support Larger File Systems and Storage

Prior to Version 5.0, there were restrictions on the support of files and file systems that exceeded 1.2 TB. This was due to the use of 32 bit fields within the stat and statfs structures. In this release, the following structures have been modified:

The changes to the stat and statfs structures are to support multiterabyte file systems. This enhancement is implemented in a way that allows current programs to continue to run without error. The stat and statfs changes are the default for programs and objects that are rebuilt on Version 5.0. The new definitions may only cause problems for applications that pass the structures as part of their API, and then only if the API and the caller were compiled on different versions of the operating system.

Any use of stat or statfs within an application or library will continue to work without error regardless of any change in the current defaults. However, if the application accepts stat or statfs structures as input to routines in their API, or returns these structures from routines in their API, a problem could occur, if the API and the code using the API were compiled on different versions of the operating system. If all the sources are compiled on systems running Tru64 UNIX Version 5.0, or higher, no problems will occur. Also if your API links with libraries and object files created on previous versions of the operating system, you can use the -D_V40_OBJ_COMPAT compiler flag to create object files that are compatible.

Applications and libraries that are built on Version 5.0 will not run on prior versions of the operating system.

8.10.3    Swap Device List and /sbin/swapdefault Moved to /etc/sysconfigtab

The list of swap devices has moved from the /etc/fstab file to the /etc/sysconfigtab file. The use of /sbin/swapdefault to indicate the swap allocation modes has been moved to the /etc/sysconfigtab file.

The swap devices and swap allocation mode are automatically placed in the /etc/sysconfigtab file during installation.

For more information, see the System Administration guide.

8.10.4    Hot Swap Support

The kernel now supports hot-swap I/O devices. This functionality provides the capability to automatically fault-in a device driver when an I/O device is hot-plugged in.

When the hardware code detects a new device is hot-plugged in and determines that the device driver is not present in the kernel, it can make a single kernel function call (for example cfg_configure) to automatically load the device's driver into the kernel. Additionally, hot-swap provides the flexibility of not having to prebuild a kernel subsystem or driver into the kernel but to dynamically fault it in at the first access of the device.

8.10.5    Kernel Attribute Changes

All kernel attribute names now use underscores. They no longer contain dashes. Old kernel attribute names using dashes are hidden; they are no longer automatically shown in the output of a query operation. Although existing tools and programs using the old attribute names still work in this release, the old names with dashes will be retired in a future release. For more information on kernel attributes, see the sys_attrs(5) reference page.

8.10.6    Kernel Tuning Attribute Changes

The default values for the maxusers and vm_page_free_target kernel tuning attributes have been modified to improve performance on larger systems. The default values set for these attributes depend on the amount of memory in the system. For more information, see the System Configuration and Tuning guide.

You can override the default values for the maxusers and vm_page_free_target attributes by setting the values in the /etc/sysconfigtab file.

Note also that the vm-mapentries, vm-vpagemax, vm-maxvas, vm-maxwire, vm-heappercent, vm-zone_size, and vm-vpagemax attributes have been removed. The vm-mapentries and vm-vpagemax attributes are now set internally to their maximum values. The vm-maxvas attribute is replaced by the max_per_proc_address_space attribure in the proc subsystem.

The vm_ubcseqstartpercent attribute is now defined to be a percentage of ubc_maxpercent instead of available memory. This allows the system to automatically adjust the sequential drain threshold when the ubc_maxpercent attribute is modified. If both the ubc_maxpercent and vm_ubcseqstartpercent attributes are specified in the /etc/sysconfigtab file, examine and adjust the vm_ubcseqstartpercent attribute according to the new specification. For example, if prior to installing Version 5.0 the ubc_maxpercent attribute is set to 40 and the vm_ubcseqstartpercent attribute is set to 20, the attributes must be set as follows to achieve the same results: 

vm:     
     ubc_maxpercent = 40
     vm_ubcseqstartpercent = 50

8.10.7    System Management

This section provides information on additional new features to the system management environment.

8.10.7.1    Enhanced Hardware Management

This release includes Enhanced Hardware Management which provides a single method for managing hardware on a Tru64 UNIX system. See the System Administration guide for more information.

8.10.7.2    Insight Manager Integrated with sys_check

The sys_check tool has been integrated with Insight Manager. An entry in the /var/spool/cron/crontabs/root file that runs sys_check once a week at 3:00 a.m. every Sunday morning and makes its output available to the Insight Manager home page. This entry is commented out by default.

You can make sys_check data available to Insight Manager by running sys_check automatically as a cron job or by running it from the SysMan Menu. In the SysMan Menu, it is listed as Create a configuration report under Support and Services.

If you want to run sys_check automatically as part of your regular system operations, remove the comment character from in front of the runsyscheck entry in the /var/spool/cron/crontabs/root file. You can edit the entry to run at a time that is convenient for your operations.

For more information, see the System Administration guide and the sys_check(8) reference page.

8.10.7.3    New /etc/rc.config.common File

Every system has a system-specific configuration file (/etc/rc.config) and a cluster-wide configuration file (/etc/rc.config.common). Both files can contain run-time configuration variables on a standalone system.

All cluster members in a cluster share the /etc/rc.config.common file. You can also create and use an /etc/rc.config.site file on multiple system for site-specific purposes.

The hierarchy of the configuration files allows an administrator to define configuration variables consistently over all systems within a LAN and within a cluster. Variables that are the same for all machines on the LAN can be defined in the site-wide file. Variables that are not defined for the site but are the same for each cluster member can be defined in the cluster-wide file. Finally, machine-specific variables can be defined in the system-specific file on each system.

The rcmgr command accesses these variables in a standard search order. The command first reads the /etc/rc.config file, then the /etc/rc.config.common file, and then the /etc/rc.config.site file, if it exists. You must manually edit the /etc/rc.config to pull in the /etc/rc.config.site.

In Tru64 UNIX Version 5.0, some run-time configuration variables that were previously stored in the /etc/rc.config file have been placed in the /etc/rc.config.common file.

You can use the rcmgr command to locate, retrieve, and set run-time configuration variables, whether they reside in the /etc/rc.config, /etc/rc.config.common, or /etc/rc.config.site file.

For more information, see the rcmgr(8) reference page.

8.10.7.4    Mail Enhancements

The sendmail program has been updated to sendmail Version 8. This version of sendmail provides enhanced functionality, including:

This release also provides the ability to connect to a POP3 server from a base UNIX client, namely MH (Rand Mail Handler). The implementation of POP3 provided with Tru64 UNIX is supplied by the MH suite of programs. This implementation also includes a POP3 client.

8.10.7.5    Enhancement to the shutdown Command

The shutdown command has been enhanced with a new option, -s, that enables you to specify that the system should execute the run-level transition scripts before reboot or halt operations. This option can be used with the -r or -h options.

8.10.7.6    Change in root crontab Behavior

The root crontab file (/usr/var/spool/cron/crontabs/root) contains three entries that clean up system log files once a week at 2:00 AM every Sunday:

One compressed backup file of each log file is retained until the next week as part of this process.

If you do not want the cron daemon to perform this cleanup or want to preserve your log files for a longer period of time, you can either change the frequency of the cleanup or remove (or comment out) the applicable entries in the root crontab file.

For more information, see the crontab(1) reference page.

8.10.7.7    NetRAIN Interface Configuration

The method for configuring a Redundant Array of Independent Network Adapters (NetRAIN) interface has changed in this release of Tru64 UNIX. If you are upgrading to Version 5.0 and you configured a NetRAIN set in an earlier version of the operating system, you will need to migrate your set to the new configuration.

Remove the current NetRAIN configuration lines from the inet.local file or rc.config file on your system, then follow the instructions in the NetRAIN section of the Network Administration guide to configure an interface.

8.10.8    AdvFS File System

The following sections describe new features for the AdvFS file system.

8.10.8.1    Fileset Creation with var Area in /usr

Selecting the "in /usr" option for the var area now creates both a usr fileset and a var fileset within the usr domain when AdvFS is selected for the usr file system. Previously, this option created a single usr fileset within the usr domain that contained both usr and var information.

This change affects users who back up AdvFS filesets with the vdump command. Previously, if your system was configured with var in the usr fileset, you needed only to back up the usr fileset. Now that this option creates a separate fileset for the var area, you must issue an additional vdump command against the var fileset or the information in the var fileset will not be backed up.

8.10.8.2    New On-Disk Format for AdvFS Domains

This release provides an improved AdvFS on-disk file system structure that obviates many of the problems with AdvFS file systems that have a large number of files. File domains created using Tru64 UNIX Version 5.0 will use the new on-disk format and will take advantage of its benefits.

File domains that were created using versions of Tru64 UNIX prior to Version 5.0 are recognized by later versions. However, domains created earlier do not support Version 5.0 enhancements. Filesets that are created in these domains, even after an operating system upgrade to Version 5.0, retain the characteristics of the older operating system. File domains created using Tru64 UNIX Version 5.0 are not recognizable by earlier versions of Tru64 UNIX and may not be used by them in any way.

For more information, see the AdvFS Administration guide.

8.10.8.3    New AdvFS System Attributes

The following system attributes have been added to AdvFS:

The AdvfsMinFreeAccess and AdvfsMaxFreeAccessPercent attributes provide you with more control over the number of available AdvFS access structures on your system. Access structures are the in-memory representation of an AdvFS file. These attributes supersede the AdvfsAccessCleanupPercent attribute, which has been removed.

The AdvfsDomainPanicLevel attribute controls what happens during an AdvFS domain panic. By default, when an AdvFS domain panic occurs, this attribute enables the system to create a crash dump without crashing the system.

Additionally, the default value of the AdvfsAccessMaxPercent attribute has changed.

For more information, see the AdvFS Administration and System Configuration and Tuning guides and the sys_attrs_advfs(5) reference page.

8.10.8.4    Ehancement to the quotacheck Command

The quotacheck -t ufs option is a new option for the quotacheck command. You can use this option in the /sbin/init.d/quota script to run the quotacheck command on only UFS file systems when quotas are enabled. For more information, see the quotacheck(8) reference page.

8.10.8.5    QUOTACHECK_CONFIG Environment Variable

The QUOTACHECK_CONFIG environment variable in the /etc/rc.config file is now available to control the behavior of the /sbin/init.d/quota script. System administrators can set the QUOTACHECK_CONFIG environment variable to specify options they want to use when running the quotacheck script during system boot. That is, you can instruct the /sbin/init.d/quota script to check both UFS and AdvFS file systems by using the QUOTACHECK_CONFIG environment variable. The default is to check the UFS file system only.

8.10.9    Logical Storage Manager Meta-Data Changes

This version of Tru64 UNIX has a new on-disk internal format for the Logical Storage Manager's (LSM) meta-data. When you boot this version of Tru64 UNIX, either during the installation process or for the first time, the LSM configuration databases for all existing LSM auto-imported diskgroups are automatically converted from the old format to the new format. Once a disk group is converted to the new format, it can not be used on earlier releases.

For LSM diskgroups that are not auto-imported (for example, a disk group that was explicitly deported), the configuration database is not automatically converted to the new format. Later when the diskgroup is imported with the voldg import command, the import will fail and cannot be used unless the administrator explicitly converts the diskgroup from the old format to the new format. To facilitate the conversion, a new LSM option to the voldg import command, -convert_old, updates a diskgroup's on-disk format.

For more information, see the Logical Storage Manager manual.

8.10.10    Development Environment

The following sections provide information on new and changed features for the Tru64 UNIX development environment.

8.10.10.1    New Features in Object-File and Symbol-Table Formats

New versions of the Tru64 UNIX object file format (Version 3.13) and symbol table format (Version 3.13) are supported in this release. These changes might impose new requirements on tools that use or modify object files.

The following list provides a brief description of the new features in Version 3.13 of the object file format:

For a full description of the new and changed features, see the supplementary documentation Tru64 UNIX Object File/Symbol Table Format Specification.

8.10.10.2    Time Zone Enhancements

Support for many new time zones has been added in this release of Tru64 UNIX. The new and updated time zone files have been added under the /etc/zoneinfo/directory. The time zone data file format has also been expanded to handle more complex transition rules. (See the tzfile(4) reference page for details). The zic compiler, zdump command, and several time-related functions in the standard C library have also been updated to support the expanded time zone data file format.

For more information, see the Installation Guide.

8.10.10.3    New Graphical Program Analysis Tools (GPA)

The Graphical Program Analysis Tools (GPA) is a set of standalone tools available from the Developers' Toolkit, which is an optional part of the Tru64 UNIX operating system available on the Associated Products Volume 1 CD-ROM The following new tools have been added:

8.10.10.4    Visual Threads Tool for Threads Debugging

Visual Threads is a new tool available on the Tru64 UNIX Associated Products Volume 2 CD-ROM that lets you analyze your multithreaded applications for potential logic and performance problems. It is licensed as part of the Developer's Toolkit for Tru64 UNIX.

You can use Visual Threads with DECthreads applications that use POSIX threads (Pthreads) and with Java applications.

Visual Threads features include the following:

8.10.10.5    OpenMP API

The cc compiler now supports the OpenMP application program interface (API). The OpenMP API supports multiplatform shared-memory programming on UNIX platforms and Microsoft Windows NT architectures. Jointly defined by a group of major computer hardware and software vendors, OpenMP is a portable, scalable model that gives shared-memory programmers a simple and flexible interface for developing parallel applications for platforms ranging from desktop to supercomputers.

The feature consists of a set of compile-time #pragma directives, a new header file (omp.h), a new runtime library (libots3.so and .a), and new command line options (-mp and -omp) that enable the pragmas and link with the new runtime library (and other existing thread libraries). If you do not specify one of the command line options to enable the directives the compiler ignores the directives (effectively treating them as comments).

8.10.10.6    Profile-Based Optimization Support

This version of Tru64 UNIX provides many compiler optimizations. For many applications, optimizations are more effective if the compiler and linker (om) have an execution profile to guide the optimizations.

The compiler and linker can use a profile to guide optimizations. In previous releases different pixie profiles were used. In this release they can use the same pixie profile. For example: 


> cc -non_shared -O3 -feedback a.out *.c
> pixie -update a.out
> cc -non_shared -O3 -feedback a.out -om *.c

These commands build the application, profile it, and rebuild it using the profile information.

The old feedback behavior is still supported, including the old semantics for the -feedback switch and the -om_ireorg_feedback option.

For more information, see the Programmer's Guide and the prof_intro(1), cc(1), pixie(1), and prof(1) reference pages.

8.10.10.7    Integrated Profiling Commands

The hiprof, pixie, third, and uprofile commands now provide complete profiling capabilities in single commands that perform all necessary instrumentation, execution, and display as selected by their command line options. The new prof_intro(1) reference page guides programmers to the most suitable tools for various tasks, such as optimizing or debugging an application.

All profilers can now profile applications that use any combination of signal handlers, and fork and pthread routines. All profilers have enhanced consistency and capabilities. For more information, see the Programmer's Guide, Programmer's Guide and the hiprof(1), pixie(1), prof(1), third(1), uprofile(1), and pfm(7) reference pages.

8.10.10.8    Extensions to atom Command's Programmable Interfaces

The atom command's IsInstType(5), GetProcInfo(5), and GetInstInfo(5) routines have new options to let tool-writers detect nonoperational (NOP) instructions and interprocedural branches.

Atom can also now instrument executables that have been optimized with the -om flag of the cc(1) command.

8.10.10.9    Change to Profiling of Threaded Programs with hiprof

The -cputime option of the hiprof(5) profiler now provides an instruction-count profile for threaded programs, the same as the -calltime option. This is due to the fact that the cpu cycles reported for kernel-threads by the RPCC instruction cannot be mapped to pthread(3) threads.

The only significant difference is that the profile is displayed as the number of instructions executed instead of cpu seconds used. The -cputime option still profiles cpu seconds for nonthreaded programs.

The new -samples option provides a profile in terms of cpu seconds for both threaded and nonthreaded programs. Therefore, you might prefer to use this option instead of the -cputime option.

The new behavior is the default for the new hiprof(1) command.

8.10.10.10    SIGFPE Signal Code Usage Changes

In previous releases, SIGFPE FAULT codes were incorrectly used instead of the correct SIGFPE TRAP codes.

A FAULT implies that the precise PC is known and that the operation can be fixed and re-executed. When the exception is imprecise and the instruction cannot be restarted, use a TRAP code.

The system software has been fixed to use TRAP codes properly. If you prefer the old behavior, you can use the configurable variable use_faulty_fpe_traps to restore the system to the old behavior, as follows: 

# sysconfig -r generic use_faulty_fpe_traps=1

If you have existing executables that cannot handle the proper TRAP codes you may want to restore the old behavior of always using FAULT codes.

Both before and after the fix, you can distinguish the true TRAP and FAULT codes by comparing the sc_pc and the sc_trap_pc fields of the signal context block. With a true FAULT, the sc_pc field contains the trigger PC of the faulting instruction. With a true TRAP, the sc_pc field contains the trap PC where the exception was realized (and so it will be the same as the sc_fp_trap_pc field). See the ieee(3) reference page for details.

To continue from a FAULT (with the default IEEE result for the exceptional operation), a signal handler might look similar to the following: 

void fpe_handler(int sig, int code, struct sigcontext *scp)
{

.
.
.
if (scp->sc_pc != scp->sc_fp_trap_pc)
scp->sc_pc += 4;

.
.
.
}

If it is a true FAULT, incrementing the sc_pc value by 4 is necessary to advance the PC beyond the exceptional instruction. (Alternatively, the handler could correct the exceptional condition and allow the faulting instruction to re-execute).

8.10.10.11    Double Long Type is Now 128 Bits

The default size of the C language long double type has changed from 64 bits to 128 bits. This allows applications to perform mathematical calculations on larger numbers and with more precision than was possible with the previous long double type. The previous long double was the same size as the double type.

8.10.10.12    64-bit Time-Handling Interfaces

A conditional set of time-related data types and interfaces has been added to support basic handling of times beyond the limits imposed by the current 32-bit time_t data type (Dec 13 20:45:52 GMT 1901 through Jan 19 03:14:07 GMT 2038).

To support customers needing access to times beyond this range, a separate 64-bit data type, time64_t, has been added along with several corresponding interfaces. This data type and all corresponding interfaces are Compaq extensions and are accessed by defining the _TIME64_T feature macro during compilation. The supporting interfaces are as follows: ctime64(), ctime64_r(), difftime64(), gettimeofday64(), gmtime64(), gmtime64_r(), localtime64(), localtime64_r(), mktime64(), settimeofday64(), and time64().

A separate timeval64 struct that contains a larger time64_t seconds field and is used by the gettimeofday64() and settimeofday64() functions is also provided.

Macros to assist developers converting between time_t and time64_t data types as well as timeval and timeval64 are also provided. These macros are: TIMET32TO64(), TIMET64TO32(), TIMEVAL32TO64(), and TIMEVAL64TO32().

While the getdate(), strptime(), and strftime() functions do not contain time_t in their external interfaces, they use the broken down tm struct, which can refer to times beyond the specified 32-bit time_t range. These modules have been rebuilt to support the larger range of times.

For more information, see the reference pages that correspond to the specific function. Also see the time(3) reference page for information on the time64_t data type, the _TIME64_T feature macro, and TIMET32TO64() and TIMET64TO32() conversion macros. For more information, see the gettimeofday(2) reference page. It contains information about the timeval64 struct and the TIMEVAL32TO64() and TIMEVAL64TO32() conversion macros.

8.10.10.13    DECthreads Process-Shared Synchronization Objects

You can now use DECthreads to create process-shared synchronization objects which protect data objects that are shared among threads running in different processes. Process-shared synchronization objects can be mutexes, condition variables, and read-write locks. This feature might facilitate the porting of some multithreaded applications to Tru64 UNIX.

The following routines support process-sychronization objects. For more information about using process-shared synchronization objects, see the Guide to DECthreads and the following reference pages:

8.10.10.14    DECthreads Debug Assistant Library

Since the release of DIGITAL UNIX Version 4.0, the Ladebug and TotalView debuggers have been able to determine the state of threads in a threaded process by using the DECthreads Debug Assistant Library. The header file for this library, pthread_debug.h, is currently available on the Tru64 UNIX kit. It provides the DECthreads debugging interface and the sole documentation regarding the interface. The libpthreaddebug library implements this interface.

The Debug Assistant Library is intended only for use by a debugger. Because it must suspend the target process, it is not intended for use by the process being debugged.

8.10.10.15    DECthreads Read-Write Locks

DECthreads support for read-write locks was added to Tru64 UNIX Version 4.0F. The following two new DECthreads read-write lock routines have been added in this release:

8.10.10.16    DECthreads Thread Stack and Backing Store Allocation

DECthreads now defines a writable stack area for each thread, based on the stack size you specify using pthread_attr_setstacksize() routine or, if you accept the default, some maximum based on uncommitted virtual memory at the time the thread is created. However, the writable stack area is made available to the thread, and the corresponding backing store is reserved, only in predefined increments as it is needed. In this way, no more backing store is reserved than the stack actually requires.

8.10.10.17    DECthreads Thread Stack Overflow Warning Area

DECthreads now configures a thread's stack area with an overflow warning area as well as a guard area. The stack overflow warning area is allocated at the overflow end of the thread's defined writable stack area and before the guard area. If the thread attempts to write in the warning area, a stack overflow exception occurs. Your program can catch this exception and continue processing as long as the thread does not attempt to write in the guard area. If the thread attempts to write in the guard area, a memory addressing violation occurs.

DECthreads configures a thread with no guard area or warning area if either of the following is true:

The size of the stack overflow warning area is platform dependent. For Tru64 UNIXVersion 5.0, and higher, the warning area is two pages (16384 bytes).

8.10.10.18    The malloc Function Is Now Tunable

The C runtime library malloc(3) function and associated functions have been modified to allow significantly better concurrency when used by multithreaded applications. Additionally, the following three new memory allocator tuning variables have been added to allow more control of allocator behavior:

As always when developing applications that make significant use of dynamically allocated memory and require maximum speed of execution, carefully read the Tuning Memory Allocation section of the malloc(3) reference page.

8.10.11    Commands and Utilities

The following sections provide information on changes to commands and utilities.

8.10.11.1    DXMtools

In this release, a graphical interface has been added to the mtools (dxmtools) utilities. The mtools utilities allow you to manipulate DOS files on a floppy diskette. You can access the dxmtools utilities from the Desktop_Apps menu in the Application Manager.

8.10.11.2    Improvements to the Program Stack

The default stack (program stack, used by the primary thread) allocation method has been changed to provide guard pages and to prevent arbitrary mmap()from interfering with the stack area.

The current stack limit is used to allocate the stack when a new program is executed. Thus, if the stack limit has been increased, the new limit will decrease the amount of virtual address space available for other purposes. Also, increasing the stack limit without increasing the address space, may result in a failure to execute the program, in which case a failure message similar to the following is displayed: 

"csh: exec failed to allocate default stack (6)"
 

If this occurs, lower the stack limit or increase the address space limit. See the reference pages for the appropriate shell for information on how to change the limits.

If the stack limit has been increased, the FreePort Express application might not run a SunOS SPARC executable file. The reason is that fpx tries to load the SPARC text at a low address, which is now occupied by the large stack. If this occurs, lower the stack limit and rerun the command.

8.10.11.3    Changes to the iostat Disk Statistics Fields

The bps and tps disk statistics fields have been widened (increased the size of the number that can be displayed) to allow for the higher numbers encountered with modern disk drives. Formerly, the bps and tp columns could merge together causing confusion about the information displayed. To allow for the wider fields and still keep the line length less than 80 columns, the default number of disks displayed was reduced from four to two.

8.10.11.4    Changes to ACL-Related Features

This release provides several enhancements to access control list (ACL) features, including the following:

8.10.12    Security

The notes in this section provide information on enhancements to security features.

8.10.12.1    Login Success and Failure Logging is Optional

In the Tru64 UNIX Enhanced Security mode, in accordance with strict C2 security rules, the time of each successful or unsuccessful login attempt is recorded in the authentication database, auth.db. This time value is the source of the last successful and unsuccessful login messages. The time value also provides Enhanced Security with the means to perform break-in detection and evasion for unsuccessful login attempts. A potential side effect of logging each successful login is frequent write activity and therefore NIS slave updates.

To allow more flexibility in Enhanced Security configuration, this version introduces selectable login logging options for the Enhanced Security model. You can now disable the logging of successful logins, unsuccessful logins, or both.

When unsuccessful login logging is disabled, break-in detection and evasion are not possible. When successful login logging is disabled, the source of the last login time is the local machine-specific utmp database rather than the authentication database. In a TruCluster environment, this means that a successful login to an account on one member is not recognized by another member, since the utmp database is member-specific.

To alter the logging options, use the new secconfig utility (available from the SysMan Security option). Under the Enhanced Security Custom option, select the Login Logging Option you want to use.

In an environment where successful logins are more frequent than unsuccessful login attempts, disabling successful login logging while enabling unsuccessful login logging will provide break-in detection and evasion while still potentially reducing write activity and NIS slave updates.

For more information, see the Security guide.

8.10.12.2    Enabling and Disabling ACLs

The way you enable and disable Access Control Lists (ACLs) has changed.

In previous versions of the operating system it was necessary to edit the configuration file, rebuild the system image, and reboot to enable or disable the use of ACLs. Now, you can enable and disable ACL processing using the Enable ACL button of the secconfig utility.

Alternatively, you can enable and disable ACL processing using the sysconfig -r sec acl_mode=enable command or it can be done automatically at system startup by adding the acl_mode entry to your sysconfig file. For more information, see the Security guide.

8.10.12.3    ACL Management Utility

This version of Tru64 UNIX provides a new ACL management utility, dxsetacl. This utility is available from the Daily Administration menu or from the /usr/bin/X11/dxsetacl file. This utility allows an appropriately privileged user to create, change, and delete ACLs. It also provides the ability to browse files and display existing ACLs. For more information, see the Security guide.

8.10.12.4    Shadow Password Support

This version of Tru64 UNIX has been modified to allow a more gradual transition from Base Security to the more strict Enhanced Security levels.

This release introduces the Shadow Passwords Only option. Enabling Shadow Passwords Only from the secconfig utility results in the same underlying security configuration changes required for eventual use of full Enhanced Security. However, this mode functions as if Base Security is still being used with the exception that account passwords are no longer found in the /etc/passwd file.

You can enable the additional Enhanced security features at any time using the secconfig, dxaccounts, or the edauth utility.

8.10.12.5    Netscape Communicator

This release contains the latest version of Netscape Communicator, Version 4.51. For information on new features included in Netscape Communicator 4.51, see the release notes at the following URL: 

http://home.netscape.com/eng/mozilla/4.5/relnotes/unix-4.51.html
 

You can also visit the Netscape Communicator New Features Tour at the following URL: 

http://home.netscape.com/communicator/v4.5/tour/index.html
 

You can also download the latest version of Netscape Communicator for Tru64 UNIX from the Netscape NetCenter's Download World Wide Web site at the following URL: 

http://home.netscape.com/download/index.html#clients

8.10.13    Window Systems Software

The following sections provide information on new and changed features for the window system software.

8.10.13.1    X Environment

The X environment has been updated to X11R6.3. This version includes the following new extensions:

Additionally, the X keyboard extension (xkb) has been updated to the Version 1.0 release rather than the prerelease version of Version 0.65 found in DIGITAL UNIX Version 4.0. The PanoramiX extension has also been moved from Advanced Development Kit (ADK) status to being a fully supported extension in this release.

8.10.13.2    Enabling Microsoft Windows Keys

The operating system now supports Microsoft Menu and Right and Left Windows keys. If you are using a PC keyboard, you can enable this feature. For more information, see the X Window System Administrator's Guide.

8.10.13.3    CDE Setup

CDE Setup enables you to set up, customize, and configure your preferences for the Common Desktop Environment (CDE) including the following:

You perform these customizations from a graphical interface; they do not require you to edit CDE and X configuration files directly. For more information, see the CDE Setup online help.

8.10.13.4    CDE Window List

The CDE window manager (dtwm) has been enhanced to provide navigation aids that enable you easily locate and access any open application. With multiple workspaces and many open windows on each workspace, it is sometimes time-consuming to search for a particular window. The CDE window manager now includes a function that displays a list of all currently active window applications. For more information, see the CDE Window List online help.

8.10.13.5    X Server Command Line Option to Get Old Vendor String

The X server has been enhanced with a new command line option, -ov, to allow applications to obtain the X server vendor string in its old format.

Beginning with this release the X server, by default, returns a vendor string of "Compaq Computer Corporation Tru64 UNIX V5.0". However, to allow for migration of applications that may have a dependency on the old vendor and product name, the -ov command line option has been added to the X server. Using this command line option causes the X server to return a vendor string of "DECWINDOWS Digital Equipment Corporation Digital UNIX V5.0".

To make use of this new option, do the following:

  1. Edit the /usr/var/X11/Xserver.conf file.

  2. Locate the line containing the -pn option at the bottom of the file and add -ov after the -pn.

You also need to stop and restart the X server using the /sbin/init.d/xlogin command for this change to take effect.

Please note that this command line option is only intended for short term usage while applications are modified to accept the new vendor string. This command line option will be removed in a future version of Tru64 UNIX.