JPS59180747A - System for automatically detecting dead lock - Google Patents

Abstract

PURPOSE:To quickly send a warning, by performing processing for detecting a dead lock with a processing routine which performs exclusive control and, processing for outputting a message indicating the occurrence of the dead lock when the dead lock is detected. CONSTITUTION:When an ENQ macro instruction is executed, an ENQ processing routine checks whether or not the resource designated by the instruction exists in the row of cue control blocks (QCB). When the resource does not exist, the routine adds a new QCB to the row and connects a cue element (QEL), and then, writes the address space identifier ASID and the address of a task control block DCB of the originating task in the QEL. When the resource exists, the routine connects a new QEL to the end of the QEL row and writes the ASID and TCB in the new QEL, and then, makes waiting display. Thereafter, the routine detects a dead lock and outputs a message indicating the incidence of the dead lock to a console picture.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention provides a deadlock detection function and a deadlock detection function in a processing routine that processes macro instructions that are intended to exclusively use sequentially available resources. The present invention relates to an automatic deadlock detection method that has a function to output a message indicating the occurrence of a deadlock on the console screen when a deadlock occurs. However, in the conventional technology, even if a deadlock occurs, the operator is not notified, which causes a delay in discovery, and it takes time to investigate the cause after an abnormality is noticed.

[Purpose of the invention]

The present invention is based on the above consideration, and an object of the present invention is to provide an automatic deadlock detection method that can detect deadlock in a timely manner and can notify an operator of the occurrence of deadlock. There is.

[Structure of the invention]

Therefore, the automatic deadlock detection method of the present invention operates when a task issues an instruction requesting exclusive control, and checks whether the resource specified by the instruction is being used by another task. , in a computer equipped with a processing routine that puts the requesting task in a waiting state if the resource is already being used by another task, causes the processing routine to perform deadlock detection processing and detects one deadlock. This feature is characterized in that a message indicating the occurrence of a deadlock is sometimes output on the console screen.

[Embodiments of the invention]

Hereinafter, the present invention will be explained with reference to the drawings. FIG. 1 is a diagram for explaining deadlock, and FIG. 2 is a diagram for explaining exclusive control of resources by the ENQ macro instruction.

As shown in FIG. 1, task T1 specifies resource A and
If the NQ macro instruction is issued, the ENQ routine is executed. The ENQ routine causes task T1 to use resource A if no other task is using resource A. When task T secures resource A, it executes subsequent processing.

When task T2 specifies resource B and issues an ENQ macro instruction, the ENQ routine is similarly executed.

If no other task is using resource B, task T can use resource B. After that, when task T issues an ENQ macro instruction specifying resource B, task T issues resource B because it is already being used by task T.

is in a waiting state. Furthermore, when task Tt issues an ENQ macro instruction specifying resource A, task T also goes into a waiting state because resource A is already being used by task T1. If tasks T1 and T2 are both in a waiting state, a deadlock occurs.

FIG. 2 explains exclusive control using the ENQ macro instruction. In Figure 2, CVT is a control vector table, QCIft queue control table,
QEL is a queue block, ASID is an address space identifier, and TCB is a task control block. In addition, the main QC
It may be considered that B and the first secondary QCB are substantially the same.

QCB is a resource unit management table and also serves as a queue terminal for QEL. QEL is a management table for each resource requester, and it shows what task in what space is using the resource specified by the QCB of the queue terminal, or what task is waiting for the resource to be used. It records what space, what task, etc. QE to QCB
If the L address is set, the resource is in use, indicating that the task recorded in the first QEL is using the resource.

Deadlock detection is performed in the following steps.

(b) When the requesting task is placed in a wait state in the processing of the ENQ macro instruction (another task has secured the requested resource), the following (b) through (d) are checked regarding the requesting task.

(b) Check whether there are resources secured by the target task (check the QEL at the beginning of each QCB corresponding to each resource, and check whether the ASID and TCB address match). Regarding all the resources we have secured,
Check (c) or d) below. If there are no resources secured by the target task, the check regarding the target task ends.

(c) Check whether the target resource is the same as the currently requested resource, and if they are the same, it is assumed that a deadlock has occurred and all checks are terminated. If they are not the same, check further).

b) Check whether there is a task waiting for the target resource (if there is a second or subsequent QEL), and if there is, check (b) or d) for all the tasks. If there is no such resource, the processing related to the target resource ends.

When an ENQ macro instruction is executed, the ENQ processing routine checks whether the resource specified by the ENQ macro instruction exists in the column of the QCB and, if not, returns the resource specified by the ENQ macro instruction. A new QCB for managing the acquired resources is connected to column 008, a QEL is connected to this new QCB, and the address space identifier ASID and TCB address of the issuing task are written in this QEL.

If so, the ENQ processing routine
Add a new Q at the end of the QEL sequence with
EL is connected, the address space identifier ASID and TCB address of the requesting task are written in this new QELO, a waiting indication is given to the corresponding TCB, and deadlock detection is then performed. When a deadlock is detected,
The ENQ processing routine performs processing to output a message on the console screen indicating that a deadlock has occurred.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, a processing routine that performs exclusive control is caused to perform processing for detecting deadlock, and when one deadlock is detected, a message indicating the occurrence of deadlock is output. Since deadlock processing is performed, the occurrence of deadlock can be quickly detected.

[Brief explanation of the drawing]

Figure 1 is a diagram to explain deadlock, Figure 2 is E
FIG. 3 is a diagram illustrating exclusive control of resources by NQ macro instructions. CVT...Control vector table, QC
B...Queue control block, QEL...
- Queue element, ASID...address space identifier, TCB...task control ambiguous block. Patent Applicant: Fujitsu Limited Representative Patent Attorney Yotsube Kyotani

Claims (1)

[Claims] Operates when a task issues an instruction that requests exclusive control, checks whether the resource specified by the instruction is being used by another task, and if the resource is already being used by another task. In a computer equipped with a processing routine that puts the requesting task in a waiting state, make the above processing routine perform deadlock detection processing and output a message indicating the occurrence of deadlock on the console screen when deadlock is detected. An automatic deadlock detection method characterized by causing processing to be performed.