HP OpenVMS Systems Documentation

OpenVMS Programming Concepts Manual

The following is a description of the callable interface to LDR$UNLOAD_IMAGE.

Unloads a removable executive image. This routine is called to unload an execlet. All resources are returned.

SYSLDR_DYN

Format

LDR$UNLOAD_IMAGE filename ,ref_handle

Arguments

filename

OpenVMS usage	character string
type	character string
access	read only
mechanism	by descriptor

The longword address of a character string descriptor containing the file name of the executive image to be unloaded. The file name must be supplied exactly as it was supplied to LDR$LOAD_IMAGE when the executive image was loaded.

ref_handle

OpenVMS usage	address
type	longword (signed)
access	read only
mechanism	by reference

The longword address of the reference handle containing the three-longword block returned by LDR$LOAD_IMAGE when the executive image was loaded.

LDR$UNLOAD_IMAGE must be called in kernel mode.

RETURNS

OpenVMS usage	cond_value
type	longword (unsigned)
access	write only
mechanism	by value

Status indicating the success or failure of the operation.

Return Values

SS$_INSFARG	LDR$UNLOAD_IMAGE was not called with two parameters.
SS$_BADPARAMS	Reference handle data inconsistent with LDRIMG block that matches the name in the first argument.
LOADER$_MARKUNL	A call was made to the LDR$UNLOAD_IMAGE routine to unload a removable executive image that already has an outstanding unload request against it.
SS$_NOPRIV	LDR$UNLOAD_IMAGE was not called in kernel mode.
SS$_NORMAL	Executive image was successfully removed from the system.
LOADER$_NOT_UNL	A call was made to LDR$UNLOAD_IMAGE to unload an executive image that is not loaded or that was not loaded with the LDR$V_UNL flag bit set.
LOADER$_UNL_PEN	A call was made to LDR$UNLOAD_IMAGE to unload an executive image that is in use. The image is marked to be unloaded later.

Description

LDR$UNLOAD_IMAGE removes an executive image from system space and returns all memory resources allocated when the image was loaded. Images can only be removed if they were originally loaded with the bit LDR$V_UNL set in the input flags to LDR$LOAD_IMAGE.

You can synchronize timed program execution in the following ways:

• Executing a program at a specific time
• Executing a program at timed intervals

The process control procedures that allow you to synchronize timed program execution require you to supply a system time value.

You can use either system services or RTL routines for obtaining and reading time. They are summarized in Table 4-10. With these routines, you can determine the system time, convert it to an external time, and pass a time back to the system. The system services use the operating system's default date format. With the RTL routines, you can use the default format or specify your own date format. However, if you are just using the time and date for program synchronization, using the operating system's default format is probably sufficient.

When using the RTL routines to change date/time formats, initialization routines are required. Refer to the OpenVMS RTL Library (LIB$) Manual for more information.

See Chapter 27 for a further discussion of using timing operations with the operating system.

To execute a program at a specified time, use LIB$SPAWN to create a process that executes a command procedure containing two commands---the DCL command WAIT and the command that invokes the desired program. Because you do not want the parent process to remain in hibernation until the process executes, execute the process concurrently.

You can also use the SYS$CREPRC system service to execute a program at a specified time. However, because a process created by SYS$CREPRC hibernates rather than terminates after executing the desired program, Compaq recommends you use the LIB$SPAWN routine unless you need a detached process.

Example 4-14 executes a program at a specified delta time. The parent program prompts the user for a delta time, equates the delta time to the symbol EXECUTE_TIME, and then creates a subprocess to execute the command procedure LATER.COM. LATER.COM uses the symbol EXECUTE_TIME as the parameter for the WAIT command. (You might also allow the user to enter an absolute time and have your program change it to a delta time by subtracting the current time from the specified time. Chapter 27 discusses time manipulation.)

! Delta time
CHARACTER*17 TIME
INTEGER LEN
! Mask for LIB$SPAWN
INTEGER*4 MASK

! Declare status and library routine
INTEGER STATUS, LIB$SPAWN

! Get delta time
STATUS = LIB$GET_INPUT (TIME,
2                       'Time (delta): ',
2                       LEN)
IF (.NOT. STATUS) CALL LIB$SIGNAL (%VAL (STATUS))
! Equate symbol to TIME
STATUS = LIB$SET_SYMBOL ('EXECUTE_TIME',
2                        TIME(1:LEN))
IF (.NOT. STATUS) CALL LIB$SIGNAL (%VAL (STATUS))
! Set the mask and call LIB$SPAWN
MASK = IBSET (MASK,0)            ! Execute subprocess concurrently
STATUS = LIB$SPAWN('@LATER',
2                  'DATA84.IN',
2                  'DATA84.RPT',
2                  MASK)

END

LATER.COM

$ WAIT 'EXECUTE_TIME'
$ RUN SYS$DRIVE0:[USER.MATH]CALC
$ DELETE/SYMBOL EXECUTE_TIME

To execute a program at timed intervals, you can use either LIB$SPAWN or SYS$CREPRC.

Using LIB$SPAWN

Using LIB$SPAWN, you can create a subprocess that executes a command procedure containing three commands: the DCL command WAIT, the command that invokes the desired program, and a GOTO command that directs control back to the WAIT command. Because you do not want the parent process to remain in hibernation until the subprocess executes, execute the subprocess concurrently. See Section 4.8.1.1 for an example of LIB$SPAWN.

Using SYS$CREPRC

Using SYS$CREPRC, create a detached process to execute a program at timed intervals as follows:

1. Create and hibernate a process---Use SYS$CREPRC to create a process that executes the desired program. Set the PRC$V_HIBER bit of the stsflg argument of the SYS$CREPRC system service to indicate that the created process should hibernate before executing the program.
2. Schedule a wakeup call for the created subprocess---Use the SYS$SCHDWK system service to specify the time at which the system should wake up the subprocess, and a time interval at which the system should repeat the wakeup call.

Example 4-15 executes a program at timed intervals. The program creates a subprocess that immediately hibernates. (The identification number of the created subprocess is returned to the parent process so that it can be passed to SYS$SCHDWK.) The system wakes up the subprocess at 6:00 a.m. on the 23rd (month and year default to system month and year) and every 10 minutes thereafter.

! SYS$CREPRC options and values
INTEGER OPTIONS
EXTERNAL PRC$V_HIBER
! ID of created subprocess
INTEGER CR_ID
! Binary times
INTEGER TIME(2),
2       INTERVAL(2)
   .
   .
   .
! Set the PRC$V_HIBER bit in the OPTIONS mask and
! create the process
OPTIONS = IBSET (OPTIONS, %LOC(PRC$V_HIBER))
STATUS = SYS$CREPRC (CR_ID,         ! PID of created process
2                    'CHECK',       ! Image
2                    ,,,,,
2                    'SLEEP',       ! Process name
2                    %VAL(4),       ! Priority
2                    ,,
2                    %VAL(OPTIONS)) ! Hibernate
IF (.NOT. STATUS) CALL LIB$SIGNAL (%VAL(STATUS))
! Translate 6:00 a.m. (absolute time) to binary
STATUS = SYS$BINTIM ('23-- 06:00:00.00', ! 6:00 a.m.
2                    TIME)
IF (.NOT. STATUS) CALL LIB$SIGNAL (%VAL(STATUS))
! Translate 10 minutes (delta time) to binary
STATUS = SYS$BINTIM ('0 :10:00.00',      ! 10 minutes
2                    INTERVAL)
IF (.NOT. STATUS) CALL LIB$SIGNAL (%VAL(STATUS))
! Schedule wakeup calls
STATUS = SYS$SCHDWK (CR_ID,         ! ID of created process
2                    ,
2                    TIME,          ! Initial wakeup time
2                    INTERVAL)      ! Repeat wakeup time
IF (.NOT. STATUS) CALL LIB$SIGNAL (%VAL(STATUS))
   .
   .
   .

When you use the system timer queue, you use the timer expiration to signal when a routine is to be executed. It allows the caller to request a timer that will activate sometime in the future. The timer is requested for the calling kernel thread. When the timer activates, the event is reported to that thread. It does not affect any other thread in the process.

For the actual signal, you can use an event flag or AST. With this method, you do not need a separate process to control program execution. However, you do use up your process's quotas for ASTs and timer queue requests.

Use the system service SYS$SETIMR to place a request in the system timer queue. The format of this service is as follows:

Specifying the Starting Time

Specify the absolute or delta time at which you want the program to begin execution using the daytim argument. Use the SYS$BINTIM system service to convert an ASCII time to the binary system format required for this argument.

Signaling Timer Expiration

Once the system has reached this time, the timer expires. To signal timer expiration, set an event flag in the efn argument or specify an AST routine to be executed in the astadr argument. Refer to Section 6.6 and Chapter 8 for more information about using event flags and ASTs.

How Timer Requests Are Identified

The reqidt argument identifies each system time request uniquely. Then, if you need to cancel a request, you can refer to each request separately.

To cancel a timer request, use the SYS$CANTIM system service.

4.9 Controlling Kernel Threads and Process Execution

You can control kernel threads and process execution in the following ways:

• Suspending a process---All kernel threads associated with the specified process are suspended.
• Hibernating a process---Only the calling kernel thread is hibernated.
• Stopping a process---All kernel threads associated with the specified process are stopped.
• Resuming a process---All kernel threads associated with the specified process are resumed.
• Exiting an image---All kernel threads associated with the specified process are exited.
• Deleting a process---All kernel threads associated with the specified process are deleted, and then the process is deleted.

There are two ways to halt the execution of a kernel thread or process temporarily:

• Hibernation---Performed by the Hibernate (SYS$HIBER) system service, which affects the calling kernel thread.
• Suspension---Performed by the Suspend Process (SYS$SUSPND) system service, which affects all of the kernel threads associated with the specified process.

The kernel thread can continue execution normally only after a corresponding Wake from Hibernation (SYS$WAKE) system service (if it is hibernating), or after a Resume Process (SYS$RESUME) system service, if it is suspended.

Suspending or hibernating a kernel thread puts it into a dormant state; the thread is not deleted.

A process in hibernation can control itself; a process in suspension requires another process to control it. Table 4-11 compares hibernating and suspended processes.

Table 4-12 summarizes the system services and routines that can place a process in or remove a process from hibernation or suspension.

The hibernate/wake mechanism provides an efficient way to prepare an image for execution and then to place it in a wait state until it is needed.

If you create a subprocess that must execute the same function repeatedly and must execute immediately when it is needed, you could use the SYS$HIBER and SYS$WAKE system services, as shown in the following example:

/* Process TAURUS */

#include <stdio.h>
#include <descrip.h>

main() {

        unsigned int status;
        $DESCRIPTOR(prcnam,"ORION");
        $DESCRIPTOR(image,"COMPUTE.EXE");

/* Create ORION */
        status = SYS$CREPRC(0,                (1)        /* Process id */
                            &image,                       /* Image */
                            0, 0, 0, 0, 0,
                            &prcnam,                      /* Process name */
                            0, 0, 0, 0);
        if ((status & 1) != 1)
                LIB$SIGNAL(status);
   .
   .
   .
/* Wake ORION */
        status = SYS$WAKE(0, &prcnam);          (2)
        if ((status & 1) != 1)
                LIB$SIGNAL(status);
   .
   .
   .
/* Wake ORION again */
        status = SYS$WAKE(0, &prcnam);
        if ((status & 1) != 1)
                LIB$SIGNAL(status);
   .
   .
   .
}

/* Process ORION and image COMPUTE */

#include <stdio.h>
#include <ssdef.h>
   .
   .
   .
sleep:
        status = SYS$HIBER();                  (3)
        if ((status & 1) != 1)
                LIB$SIGNAL(status);
   .
   .
   .
        goto sleep;
}