JPH036654A - Mutual exclusion system - Google Patents

Abstract

PURPOSE:To attain the effective mutual exclusion control in a system where a 'dangerous part' is frequently carried out by deciding selectively a method to obtain a process in accordance with the state of a process which owns the executing right of the 'dangerous part' and performing selectively the process switching and error processes. CONSTITUTION:The processors 1 and 2 and a main storage 3 are connected to a system bus 4, and each processor can read and write the optional data stored in the storage 3 and also can execute an optional program. The storage 3 stores a processor control block, a processor waiting process matrix, a key, a 'dangerous part', etc. When plural processes carry out the inlet operations at one time, only a single process can carry out the 'dangerous part'. Thus other process are unable to carry out the 'dangerous part' before the process finishes an outlet operation. As a result, the effective mutual exclusion control is attained in a system where a user process having the small process value carries out frequently the 'dangerous part'.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a mutual exclusion scheme for shared resources in a multiprocessor system.

(Prior Art) As a basic method of mutual exclusion for shared resources in a multiprocessor system, the spin lock method and the
There is a binary semaphore method. The configuration and drawbacks of each method are described below.

(a) Spin-lock system (a-1) Structure of spin-lock system FIG. 1 shows an example of the structure of a system that performs mutual exclusion using the spin-lock system. The processor 1.2 and the main storage device 3 are connected to one system bus 4.

Each processor is capable of reading and writing arbitrary data stored in the main memory 3 and executing arbitrary programs. As shown in the figure, the main memory 3 stores keys, critical parts, 1 ock operations, un 1 ock operations, etc.
Operation programs, etc. are stored. What is the dangerous part?
A program that can only run one process at a time and is subject to mutual exclusion.

The value of the key is 1 or 0, and indicates whether a process exists that has the right to execute the "sensitive part." 1
The ock operation and the unlock operation are programs that mutually exclude "critical parts".

FIG. 2 shows an operation procedure for performing mutual exclusion using a 1 ock operation and an un 1 ock operation. If multiple processes execute the 1ock operation at the same time, the functionality of the 1ock operation described below allows only one process to execute the "tight part". This process is unlocked
Until the k operation is performed, other processes remain in the "critical part"
I idled and waited without entering it.

FIG. 3 is an example of realizing the 1ock operation. The 1Ock operation is performed in the following steps.

■Read out the key. Let the read value be X. Next, put the key 1
Make it. Here, reading and writing to the key is done by te.
Memory reading and writing are performed by undivided basic operations such as st-and-set instructions.

(2) If X is O, the 1ock operation is completed.

If not, go to ■ and idle.

FIG. 4 is an example of implementation of the unlock operation.

The unlock operation writes 0 to the key. uniock
When the operation writes 0 to the key, the process that was idling during the 1ock operation reads 0 as the key value and executes the 1ock operation.
You will be able to finish the operation and move on to the "dangerous part".

(a-2) Disadvantages of spin lock method Since the process idles, checks the state of the key, and waits until the key becomes O, the following restrictions are imposed on the "critical part" of the program.

・It must be a short program ・It must be possible to avoid “dangerous parts” by disabling interrupts, etc.
To prevent the processor from being taken away in the middle of execution.

- Do not perform operations such as input/output that result in a wait state.

- The processor cannot be taken over by the exception handling program due to an error.

Due to the above limitations, the spinlock method is generally used for mutual exclusion of limited programs such as an OS kernel, and has the disadvantage that it is not suitable for normal user programs. That is, when the spin lock method is used in a user program, the following problem occurs.

- Due to a bug, there is a possibility that interrupts will be disabled for a long time without leaving the "critical part".

■The 1ock operation does not have a function to detect a change in the state of the process that owns the execution right for the "critical part". Therefore, when a process that has the right to execute a "critical part" enters a wait state or has its processor taken away, the processor running the process that continues to idle will continue to idle, wasting a lot of time. .

■A simple mistake in how to use the 1ock operation can easily lead to a deadlock.

However, the 1ock operation does not have a function to check which process owns the execution right for the "sensitive part", so it is possible that a simple program error may result in the user never being able to obtain the execution right for the "sensitive part". It is not possible to detect a waiting state and perform appropriate processing.

- This conflicts with the general goal of O8, which is to have user processes run as long as possible without having to worry about taking away the processor.

(b) Binary semaphore system (b-1) Configuration of binary semaphore system FIG. 5 shows an example of the configuration of a system that performs mutual exclusion using the binary semaphore system. The processor 1.2 and the main storage device 3 are connected to one system bus 4.

Each processor is capable of reading and writing arbitrary data stored in the main memory 3 and executing arbitrary programs. As shown in the figure, the main storage device 3 stores a semaphore S, a waiting process queue for the semaphore S, a processor waiting process queue, a P operation, a ■ operation, and a process switching program, in addition to the critical parts described above.

As mentioned above, the critical parts are programs that can only execute one process at a time, and are subject to mutual exclusion. Semaphore S takes 0 or 1,
Each "sensitive part" is in use, or
Indicates that it is vacant. The semaphore S waiting process queue is O8 control data in which processes waiting for the use of the "critical part" to finish are lined up.

The processor queue is O8 control data that lines up processes that can be executed as soon as a processor is allocated. P operation and V operation are programs that mutually exclude "critical parts".

FIG. 6 shows an operation procedure for mutual exclusion using the P operation and the V operation. If multiple processes execute a P operation at the same time, the functionality of the P operation described below allows only one process to execute the "critical part." Until this process performs the V operation, other processes are kept waiting in the semaphore S waiting process queue.

FIG. 7 is an example of realizing the P operation. The P operation is performed in the following steps.

■Read semaphore S. The read value is set as old. Next, if old is 1, S is subtracted by 1 and becomes 0.

Here, reading and writing to a semaphore are visible to the process as basic operations that are not subdivided.

(2) If old is 1, end the P operation. If not, the process executing the P operation is added to the end of the semaphore waiting process queue, and control is transferred to the process switching program.

FIG. 8 is an example of realizing the operation (2). ■The operation is to check whether the semaphore waiting process queue is empty, and if it is empty, send 1 to semaphore S.
Write and end the operation. If it is not empty, the first process is taken out of the semaphore S queue, placed at the beginning of the processor waiting process queue, and the operation is completed.

When the process switching program is executed after an appropriate period of time has elapsed, the process is taken out from the processor waiting process queue, and processing is resumed from the point at which the P operation ended.
Perform the “dangerous part”.

(b-2) Disadvantages of the binary semaphore method In the binary semaphore method, when multiple processes try to execute the same "critical part", for all processes that cannot be decremented by 1 with the P operation, the following Such operations are carried out, and there is a disadvantage that process switching occurs frequently in mutual exclusion.

・Every time a P operation is performed, a process is placed in the waiting process queue of the semaphore S, and process switching is performed. ・Whenever a V operation is performed, a process is moved from the waiting process queue of the semaphore S to the waiting process queue for processor allocation. ■
After the operation, a process switch occurs.

The amount of processing required for the above operation is extremely large compared to the amount of processing required for mutual exclusion using the spinlock method, and the above operation is essential if there is already a process that owns the execution right for the "sensitive part". , it has the disadvantage that it appears as a large overhead in processing the process. In particular, the binary semaphore method has the disadvantage that in systems where processes with a small amount of processing frequently execute "critical parts", the overhead for mutual exclusion becomes extremely large in the processing of the entire process. In addition, since the P operation does not have a function to read the state of the process that owns the execution right of the "critical part," it is difficult to detect and appropriately handle errors such as deadlocks that occur due to the ownership relationship of the execution right. The drawback is that it cannot be done.

(Problems to be Solved by the Invention) The problems to be solved by the present invention are the drawbacks that the spinlock system is unsuitable for mutual exclusion of user programs due to large restrictions imposed on programs in "critical parts"; The disadvantage of the binary semaphore method is that there is a large overhead due to frequent switching of processes due to mutual exclusion, and in both methods, the mutual exclusion means has the ability to read the state of the process that owns the execution right of the "critical part". The disadvantage is that it is not possible to detect and process errors in programs related to mutual exclusion because the
It is an object of the present invention to provide an efficient mutual exclusion scheme in systems where low-volume user processes frequently perform "tight parts."

[Structure of the invention]

(Means for Solving the Problems) The present invention provides a function to read the execution state of a process that owns the execution right of a "critical part" and a function to detect that the execution right cannot be newly obtained. , and by providing the mutual exclusion means with the ability to control the execution of processes waiting to obtain execution rights.
Depending on the state of the process that owns the execution right for the "critical part," the process that is waiting to get the execution right may be made to wait with a processor allocated to it, take away the processor, or perform erroneous processing. The most important feature is mutual exclusion by controlling execution to ensure

(Operation) According to the above-described mutual exclusion method of the present invention, a method for making a process wait is selectively determined according to the state of the process that owns the execution right for the "critical part", and a process switching,
Then, error processing is selectively performed.

(Example) Hereinafter, an example of the present invention will be described in detail using the drawings.

FIG. 9 is a system configuration diagram of an embodiment of the present invention.

Processor 1.2 and main memory 3 are connected to one system bus 4. Each processor reads and writes arbitrary data stored in the main storage device 3,
And it is possible to execute any program. As shown in the figure, the main storage device 3 includes process control blocks,
Processor waiting process queues, keys, critical parts, entry operations, exit operations, exception handling programs, and process switching programs are stored.

The process control block (PCB) is control data created by the O8 for each process, and the execution status of the process, the allocation of resources to the process, etc. are written therein. The processor waiting process queue is O8 control data in which the PCBs of processes that can be executed immediately upon allocation of a processor are arranged. The key consists of a PCB address and a lock variable. The PCB address stores the start address of the PCB of the process that has the right to execute the "critical part" or 0. The lock variable is
Indicates whether there is a process that has the right to execute the "sensitive part". The critical parts are programs that can only have one process running at a time, and are subject to mutual exclusion. Entry operations and exit operations are programs that mutually exclude "tight parts." The exception handling program is a program that handles errors related to mutual exclusion, and the process switching program is a program that allocates processors to processes.

FIG. 10 is an operational procedure for performing mutual exclusion using an inlet operation and an outlet operation. When multiple processes perform population operations at the same time, the population operations feature described below allows only one process to perform the "sensitive parts." No other process can enter the "critical part" until this process performs the exit operation.

FIG. 11 is an example of an entrance operation. In accordance with FIG. 11, an implementation example of the entrance operation will be described in detail below. key P
The initial values of the CB address and lock variable are set to 0, and the process executing the entry operation is called process A.

■Read the PCB address of the key. Let the read value be a. If a is O, go to ■. a is process A (
In the case of the start address of the PCB of the current process), it means that there is an error in the program of process A and that the right to execute the "critical part" cannot be newly obtained, so control is transferred to the exception handling program. In other cases, go to ■.

■Read the lock variable of the key. Let the read value be X. Next, set the key's lock variable to 1. Here, reading and writing to the lock variable is done using test-and-se.
This is performed by an atomic operation using a t instruction or the like. If X is 0, P of the process performing the entry operation
Write the first address of the CB to the PCB address of the key,
Finish the entrance operation. If X is not 0, go to ■.

(2) Read the process status field S from the process control block whose starting address is a. If the value of S indicates that the process is running, go to ■. If the value of S indicates that the process has terminated, then control is transferred to the exception handling program, since it means that the execution right for the "critical part" will never be returned. In other cases, the PCB of process A is placed at the end of the processor waiting process queue, and control is transferred to the process switching program.

FIG. 12 is an example of an implementation of the exit operation. The exit operation first sets the key's PCB address to 0, then sets the key's lock variable to 0.

〔Effect of the invention〕

As explained above, the present invention provides a function to read the execution state of a process that owns the execution right of a "critical part", a function to detect that the execution right cannot be newly acquired, and Mutual exclusion means is equipped with a function to control the execution of processes that are waiting to obtain the execution right, so that depending on the state of the process that owns the execution right of the "critical part", the process that is waiting to obtain the execution right is Execution is controlled so that a process with a small amount of processing capacity is forced to wait with a processor allocated to it, takes away the processor, or detects an error and performs exception handling. Efficient mutual exclusion control can be realized in a system that executes "parts".

Furthermore, in the mutual exclusion method of the present invention, when the process that owns the execution right for the "critical part" is in the execution state, there is no write instruction to memory in the procedure in which other processes idle and wait. There is. This means that in the case of a shared memory multiprocessor with a cache, in the above state, data in the cache memory is only referenced, and access to the main memory or data transfer to the cache of other processors is not possible. This means that the performance of other processors is not degraded due to idling, which is an advantage.

[Brief explanation of the drawing]

Figures 1 to 4 are drawings related to the spinlock method, which is a conventional mutual exclusion technology, and Figure 1 is a system configuration diagram of an embodiment of the spinlock method, and Figure 2 is a system configuration diagram using the 1lock operation and the unlock operation. FIG. 3 is a flowchart of a detailed procedure for a 1lock operation, and FIG. 4 is a flowchart of a detailed procedure for an unlock operation. Figures 5 to 8 are diagrams related to the binary semaphore system, which is a conventional mutual exclusion technique. Figure 5 is a system configuration diagram of an embodiment of the binary semaphore system, and Figure 6 shows the P operation and V operation. Flowchart of operating procedures for mutual exclusion using
The figure is a flowchart of the detailed procedure of the P operation, and FIG. 8 is the flowchart of the detailed procedure of the V operation. 9 to 12 are drawings related to an embodiment of the present invention, and FIG. 9 is a system configuration diagram of an embodiment of the present invention.
Figure 0 is a flowchart of the operation procedure for mutual exclusion using the inlet operation and outlet operation, which are the mutual exclusion means of the present invention.
FIG. 1 is a flowchart of the detailed procedure of the entrance operation, and FIG. 12 is a flowchart of the detailed procedure of the exit operation.

Claims (1)

[Claims] Mutual exclusion of the means by which multiple processes running on different processors can share resources and the right to start execution (called the execution right) of a part that can only be used by one process at a time (called the critical part) In a multiprocessor system having means for The processor has a function to control the execution of a process that is waiting to obtain the execution right, and according to the state of the process that has the execution right of the "critical part", the processor A mutual exclusion method that is characterized by controlling execution so that a processor remains allocated and waits, a processor is taken away, or an error process is performed.