JPH09282182A - Pre-emptive control method for computer system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve efficiency for developing an operating system by enabling the nest control of pre-emptive by inhibiting the interruption of a task under execution when the value of a counter is positive and enabling the interruption of the task under execution correspondently when the value of the counter becomes '0'. SOLUTION: Concerning a computer system for parallelly executing plural tasks, when a function for requesting the inhibition of execution stop for the task under execution is issued from an application or a system call routine, a counter 12 showing the possibility/impossibility of execution stop for the task under execution is incremented. When the value of the counter 12 is positive, a real scheduling operation is delayed until the inhibition of execution stop for the task under execution is canceled. Further, when the value of the counter 12 is '0' and a scheduling request to the other task is generated, the scheduling operation is immediately performed. Through such control, the nest control of pre-emptive is enabled.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a preemption control method in a computer system, and more particularly to a preemption control method in a computer system suitable for an operating system providing a multitask environment.

[0002]

2. Description of the Related Art An operating system is provided with an interface called an "system call" with an application in order to provide its function to an application program (hereinafter referred to simply as "application"). When an application issues a system call, a routine inside the operating system (system call routine) is started. By the way, while not only the application but also the system call routine is being executed, the execution of the task is interrupted due to the CPU time allotted to the task being executed being used up, and the routine driven by the CPU at certain time intervals. May issue a scheduling request for another task. When the task execution is temporarily suspended in this way, the task being executed is said to be "preempted".

An application or system call routine has a piece of code called the danger zone, and while one task is running in the danger zone, no other task is allowed to run in the danger zone. (Co-authored by Peterson / Silver Shatz, co-translated by Koichi Utsumiya / Akira Fukuda, "The Concept of Operating Systems", Volume 1, Baifukan, page 314, line 16 to page 319, line 31.
987). If a request for scheduling another task is issued while a running task is running in the dangerous area, if another task is scheduled immediately, a new task will be created even if there is a task running in the dangerous area. Scheduled tasks may run in the danger zone.

Conventionally, as a method of limiting the upper limit of the number of tasks that run in a dangerous area to 1, a method using "preemption control" is known. For example, μITRON3.0 provides the following interfaces for performing preemption control. (Supervised by Ken Sakamura, "μITRON
3.0 Standard Handbook ", Personal Media, 74
(Page to page 77, 1993) (1) dis_dsp: transitions a running task to a preemption prohibited state (described later).

(2) ena_dsp: Transitions the running task to the preemptible state (described later).

When an application or system call routine issues the dis_dsp function, then ena
Until the _dsp function is issued, the running task is in the "preemption disabled state", that is, the running task transits to a state where it is not preempted even if a scheduling request for another task is issued. dis_dsp function and ena_d
The sp functions are always called in pairs.

When the application or system call routine issues the ena_dsp function, the task under execution is in the "preemptable state", that is, if a scheduling request for another task is issued, the task under execution is immediately preempted. To return to the normal state. If there is a scheduling request for another task issued while the running task is in the preemption disabled state, the other task is scheduled when this ena_dsp function is issued.

The "preemption control" is to change the preemption state (general term of "preemption prohibited state" and "preemption enabled state") of the task under execution.

In order to limit the upper limit of the number of tasks running in the dangerous area by the preemption control, the running task is transited to the preemption prohibition state immediately before entering the dangerous area, and the preemption is performed immediately after leaving the dangerous area. All you have to do is return to the enabled state.

In this way, the task is placed in the preemption prohibited state while the task is running in the dangerous area. Therefore, it is not possible that another task is scheduled while one task is running in the dangerous area and the scheduled task enters the dangerous area.

[0011]

In order to limit the upper limit of the number of tasks that run in a dangerous area using preempt control, a preempt control function is embedded in the system call routine. Therefore, a function (ena_dsp function) that transitions the running task to the preemptible state
Is also embedded in the system call routine.

Now, consider a case where another system call routine is called from within the danger area of the system call routine when the task under execution is in the preemption disabled state. In the called system call routine,
It is assumed that a function that changes the running task to the preemptible state is used.

In this case, even though the task being executed is transited to the preemption disabled state before this system call routine is called, the task being executed is transited to the preemptible state in the called system call routine. Since the function to issue is issued, the running task returns to the preemptible state. That is, the task under execution becomes preemptible even though it is in the dangerous area, and the program malfunctions.

Therefore, when calling another system call routine from within the dangerous area, it is necessary to separately prepare a routine excluding the preemption control function from the system call routine to be called, and call that routine. There was a problem that the development efficiency of the system was reduced.

A similar problem occurs when a library routine (a shared routine used by a plurality of system call routines) that uses a function for shifting a running task to a preemptable state is called from within the dangerous area. Again, the operating system development efficiency is reduced.

A system call routine or a library routine that uses a function for transitioning a task under execution to a preemptible state even from inside the dangerous area (hereinafter, these two are collectively referred to as a system call routine) In order to be able to call, a preemption control method capable of nesting a pair of functions that transit the preemption state is required.

An object of the present invention is to provide a nestable preemption control method.

[0018]

According to the present invention, in a computer system which executes a plurality of tasks in parallel, when a function requesting prohibition of interruption of execution of a task under execution is issued from an application or a system call routine. Then, the counter indicating whether or not the execution of the task under execution can be interrupted is incremented. Also, when a function requesting cancellation of the suspension of execution of a running task is issued from an application or a system call routine, the counter is decremented. If the value of the counter is positive, the actual scheduling operation is delayed until the suspension of execution of the executing task is released even if a scheduling request is issued to another task. Furthermore, if the value of the counter is 0, the scheduling operation is immediately performed when a scheduling request to another task is generated.

The above control enables preempt nest control.

[0020]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a schematic diagram of a module group and data necessary for implementing the present invention.

In FIG. 1, 11 indicates a preemption prohibition flag. The operating system changes the operation of the scheduler 14 by turning ON / OFF the preemption prohibition flag 11 and transits the preemption state.
When the flag is turned ON, the preemption state of the task under execution is in the preemption prohibited state, and when it is OFF, it is in the preemption enabled state.

The flow chart of the scheduler 14 is shown in FIG.
Shown in

The scheduler 14 starts with step 41.
In, the preemption prohibition flag 11 is checked. When the preemption prohibition flag 11 is ON, it indicates that the running task is in the preemption prohibition state. Therefore,
When the scheduling request flag 13 is turned ON in step 42, the scheduler 14 immediately returns to the routine that called the scheduler (for example, the system call routine 17).

In step 41, if the preemption prohibition flag 11 is OFF, the task under execution is in the preemption enabled state. Therefore, the scheduler performs the normal scheduling operation in step 43 and returns to the routine that called the scheduler.

The counter 12 is a counter for holding the number of times of acceptance of the preemption prohibition request incorporated according to the present invention. The preemption prohibition request issued from the system call routine 17 is processed by the preemption prohibition function 15, in which the counter 12 is also incremented. On the other hand, the request for canceling the preemption prohibition is processed in the preemption prohibition cancellation function 16, in which the counter 12 is decremented. This will be described later in detail.

When the value of the counter 12 is positive, it indicates that the number of times the preemption prohibition function 15 is issued is greater than the number of times the preemption prohibition cancel function 16 is issued, so that the task being executed cannot enter the preemption prohibition state. I have to. Similarly, when the value of the counter 12 is 0, the task being executed must be in the preemptable state. Therefore, when the value of the counter 12 is positive, the preemption prohibition flag 11 is turned on, and when the value of the counter is 0, the preemption prohibition flag 11 is turned off.
5 and the preemption prohibition cancellation function 16 are realized. This will also be described later in detail.

The system call routine 17 needs to transit the preemption state of the task being executed to the preemption prohibited state or the preemption enabled state in order to realize the dangerous area. In that case, the following interfaces are used to call the preemption prohibition function 15 and the preemption prohibition release function 16.

<Function name> preempt_disable (* level) <argument> * level: This function and function preempt_enab
Specifies the address that receives the depth of the nest of pairs formed by le.

<Description> This function shifts the task under execution to the preemption disabled state. After issuing this function and transitioning to the preemption disabled state, the function preempt_en
This function and the function preempt_enable are in the interval from issuing the "able" to returning to the preemptible state.
You may issue e as a pair. That is, the pair of this function and the function preempt_enable can be nested. This function and function preem issued inside the nest
pt_enable does not make a state transition between the preemption prohibited state and the preemption enabled state. * Le
The depth of this nest is returned to the memory area indicated by vel.

<Function name> preempt_enable (level) <argument> level: function preempt_d which is a pair with this function
Specifies the nesting level returned by isable.

<Description> This function returns the running task to the preemptible state. Function preempt_di
The function preempt_disable and this function may be issued as a pair in the interval from issuing the "save" command to transit to the preemption prohibited state and then issuing this function and returning to the "preemption enabled" state. That is, the function pr
The pair of eempt_disable and this function can be nested. Function preempt issued inside the nest
_Disable and this function do not perform the state transition between the preemption prohibited state and the preemption enabled state.
In the level, the depth of this nest, that is, the value of the level obtained by issuing the paired function preempt_disable is specified. The kernel also holds this nest depth, and if it does not match the nest depth specified by the argument, the system call routine 17
Notify the error.

When the system call routine 17 uses the preemption control function, first, when the preemption prohibition function 15 (preempt_disable function) is called, the current nesting depth returned by the preemption disable function 15 is determined by the function itself. Stored in the work area 18. Then, the preemption prohibition release function 16 (preempt_ena)
When calling the ble function), the value stored in the work area 18 is specified as an argument of the function. The preemption prohibition canceling function 16 compares the value of this argument with the value of the counter 12 and, if they do not match, notifies the system call routine 17 of an error.
premp to be paired with the mpt_enable function
It is possible to easily detect a program error that the t_disable function does not exist.

The processing procedure of the preempt_disable function will be described with reference to FIG. 2, and preempt_ena will be described.
The processing procedure of the ble function will be described with reference to FIG.

In the preempt_disable function, first, in step 21, the value of the counter 12 is incremented. Then, in step 22,
The value of the counter 12 after the increment is copied to the address given by the first argument. Then, in step 23, the preemption prohibition flag 11 is turned on and the process returns.

In the preempt_enable function,
First, in step 31, referring to the counter 12,
Compare with the value of the first argument. As a result, if they match,
In step 32, the value of the counter 12 is decremented. If they do not match, system call routine 1
Notify 7 of the error. Next, in step 33, the value of the counter 12 after decrement is checked. At this time, if the value of the counter 12 is not 0, the process ends normally. Also, 0
If so, in step 34, the preemption prohibition flag is turned off. Further, in step 35, the scheduling request flag 13 is checked, and if it is ON, the scheduling request flag 13 is turned OFF in step 36, and then the scheduler is activated in step 37 to normally end. If the scheduling request flag 13 is OFF in step 35, steps 36 to 37 are not performed.

Preempt_disa as described above
If the ble and preemptt_enable functions are used, the issued p
pr the same number of times as the reempt_disable function
Unless the empty_enable function is issued, the running task does not return to the preemptable state. Therefore, even from inside the dangerous area (when the running task is in the preemption disabled state), it becomes possible to call a system call routine that uses a function for transitioning the running task to the preemptable state.

[0038]

As described above, according to the present invention, a pair of a preemption prohibition request and a preemption prohibition cancellation request can be issued in a nested manner. Then, unless the preemption prohibition cancel function is issued the same number of times as the issued preemption prohibition function, the task under execution does not return to the preemption enabled state. Therefore, even if the running task is in the preempt disabled state, a system call that uses a function that transitions the running task to the preempt enabled state.
Routines can be called, improving the operating system development efficiency.

Even if the preemption prohibition request and the preemption prohibition cancellation request are not correctly paired, the fact can be detected by the error return of the preemption prohibition cancellation function, so that the reliability of the operating system is improved. Can be increased.

[Brief description of drawings]

FIG. 1 is a schematic diagram of a system to which the present invention is applied.

FIG. 2 is a flowchart of processing for prohibiting preemption.

FIG. 3 is a flowchart of processing for canceling preemption prohibition.

FIG. 4 is a flowchart of the scheduler.

[Explanation of symbols]

11: preemption prohibition flag, 12: counter, 1
3: Scheduling request flag

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Takahiro Nakano 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture Inside the Hitachi, Ltd.System Development Laboratory (72) Inventor Kazuto Serizawa 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture Stock Company Hitachi, Ltd. System Development Laboratory

Claims (3)

[Claims] 1. A preemption control method in a computer system for executing a plurality of tasks in parallel, wherein a function for requesting prohibition of interruption of a running task is executed by an application corresponding to the task or an operating system related to the execution of the task. In response to the issuance of a routine inside the system, a function that increments the counter that indicates the number of nests for which suspension of a running task is prohibited and requests cancellation of suspension of the running task is executed by the application or The counter is decremented in response to the issuance of a routine inside the operating system related to the execution of the task. When the value of the counter is positive, interruption of the task under execution is prohibited, and the value of the counter is 0. It becomes possible to interrupt the running task depending on A preemption control method in a computer system, comprising: 2. The preemption control method for a computer system according to claim 1, wherein the value of the counter is determined in response to a task scheduling request, and if the counter value is positive, A preemption control method in a computer system, wherein a scheduling operation is delayed until a task under execution can be interrupted, and when the value of the counter is 0, the scheduling operation is immediately performed. 3. The preemption control method in a computer system according to claim 1, wherein the updated value of the counter is returned as return information of a function requesting prohibition of interruption of an executing task, and Specify the number of nests of suspending prohibition of the running task in the argument of the function that requests cancellation of suspending prohibition of the task. If the above argument and the value of the above counter do not match, A preemption control method in a computer system, characterized in that an error is notified to an application that has issued a function requesting cancellation of suspension inhibition or a routine inside the operating system related to execution of the task.