JP2009217750A - Computer system, input/output priority control method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To cope with a problem that input-output processing by a real-time OS is prevented by input-output processing by a general purpose OS, and consequently a response of the real-time OS lowers. <P>SOLUTION: A computer system has a sharing device 6 shared by the real-time OS 3 and the general purpose OS 5, a priority controller 32 which divides a first input/output request issued to the versatile OS 5 into two or more second input/output requests which demand that an input-output block to which input/output is required by the first input/output request, should be divided into two or more partial blocks for input/output, and an input-output processing interface 36 which performs input-output processing to the sharing device 6 in response to the second input/output requests and the third input/output request issued to the real-time OS 3. The priority controller 32 responds to the properties of an issuance source of the first input/output request corresponding to two or more second input/output requests and the issuance source of the third input/output request, and performs priority control over input-output processing to the sharing device 6. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a computer system, an input / output priority control method, and a program, and more particularly to conflict control of input / output processing in a hybrid OS in which a real-time OS and a general-purpose OS are incorporated.

  For embedded systems such as car navigation, digital cameras, information appliances, mobile phones, and other built-in systems, it is necessary to increase the functionality of the system by performing multimedia processing that handles audio data and video data in addition to the real-time processing that has been performed conventionally. It has been.

  One method for realizing high functionality is to use a hybrid OS equipped with two types of OSs, a real-time OS (operating system) that performs real-time processing and a general-purpose OS that performs multimedia processing (for example, JP 2005-234658 A). By adopting such a hybrid OS, it is possible to realize a system having both features of the two OS, such as high real-time property and high reliability possessed by the real-time OS, and high scalability and high throughput possessed by the general-purpose OS.

  When hybridizing an OS in an embedded system, it is common to share a disk between the OSs. This is because it is difficult to provide a disk for each OS due to space-saving restrictions.

  However, when a shared disk is provided, the input / output processing of the real-time OS may be hindered by the input / output processing of the general-purpose OS, and the response of the real-time OS may decrease. One factor in the decrease in response of the real-time OS is that the general-purpose OS tends to have a large input / output block size for input / output requests in order to improve throughput. If the input / output block size of the input / output request of the general-purpose OS is large, the execution time of the input / output processing per one input / output request becomes long, and the input / output processing requested from the real-time OS cannot be started and delayed. .

  On the other hand, input / output processing directed only at improving the response of the real-time OS may not be preferable. For example, a task that does not require real-time performance may be executed in the real-time OS, and a process that requires a certain level of real-time performance may be executed in the general-purpose OS. In such a case, it is not preferable that the process executed in the general-purpose OS is delayed by the input / output processing of the real-time OS.

  Against this background, while optimizing the input / output processing of the real-time OS, the competition control of the input / output processing of the entire system is optimized while dealing with the problem that the response of the real-time OS is reduced by the input / output processing of the general-purpose OS. There is a need to provide technology to do this.

  Relatedly, Japanese Patent Laid-Open No. 11-39246 discloses competition control for input / output processing. More specifically, this publication discloses that the disk driver is provided with a normal priority I / O request queue and a high priority I / O request queue. For I / O requests requested with normal priority, I / O processing is performed on the disk in units of I / O sub-blocks obtained by dividing the I / O block. When a high-priority I / O request occurs, the input / output processing for the I / O request is interrupted at the break of the I / O sub-block and processed first.

  Japanese Patent Application Laid-Open No. 10-49388 also discloses contention control for computer input / output processing. This publication determines the total data size at the time of locking to be transferred in input / output processing during lock setting based on the priority of the process, the transfer data size of the input / output device and the input / output data size required by the process, and It is disclosed that a large total data size at the time of locking is allocated to a process with high priority.

However, these publications have no suggestion regarding control of input / output processing to a shared disk in a hybrid OS in which a real-time OS and a general-purpose OS are incorporated.
JP 2005-234658 A JP-A-11-39246 Japanese Patent Laid-Open No. 10-49388

  The object of the present invention is to optimize the competition control of input / output processing as a whole system while dealing with the problem that the input / output processing of the real-time OS is obstructed by the input / output processing of the general-purpose OS and the response of the real-time OS decreases. It is to provide a technique for performing.

  In one aspect of the present invention, a computer system in which a hybrid OS composed of a real-time OS and a general-purpose OS is installed is issued to the shared device shared by the real-time OS and the general-purpose OS and the general-purpose OS. The first input / output request is divided into a plurality of second input / output requests for requesting to input / output by dividing the input / output block requested for input / output by the first input / output request into a plurality of partial blocks. Input / output request dividing means for performing input / output processing to the shared device in response to the plurality of second input / output requests and a third input / output request issued to the real-time OS; It has. The input / output processing means is configured to input the input to the shared device according to a property of an issuer of the first input / output request and an issuer of the third input / output request corresponding to the plurality of second input / output requests. Controls the priority of output processing.

  According to the present invention, the competition control of the input / output processing as the whole system is optimized while dealing with the problem that the input / output processing of the real-time OS is obstructed by the input / output processing of the general-purpose OS and the response of the real-time OS is reduced. Techniques for doing so are provided.

  FIG. 1 shows a system configuration diagram of an embedded system 10 which is an embodiment of a computer system of the present invention. In the embedded system 10, a hybrid OS including a real-time OS 3 and a general-purpose OS 5 and an inter-OS communication driver 4 that performs inter-OS communication between the real-time OS 3 and the general-purpose OS 5 are installed. The real-time OS 3, the inter-OS communication driver 4, and the general-purpose OS 5 are all software programs that operate on the embedded system 10, and use hardware resources (an arithmetic device, an input / output device, a storage device, etc.) of the embedded system 10. Executed. The embedded system 10 includes a shared disk 6 that is shared between the real-time OS 3 and the general-purpose OS 5.

  The embedded system 10 is further installed with an application 1 that operates on the real-time OS 3 and an application 2 that operates on the general-purpose OS 5. The application 1 that operates on the real-time OS 3 executes various tasks, and similarly, the application 2 that operates on the general-purpose OS 5 executes various processes. In FIG. 1, tasks executed by application 1 are indicated by reference numerals 11 and 12, and processes executed by application 2 are indicated by reference numerals 21 and 22.

  The general-purpose OS 5 includes an input / output acceptance interface 51, and accepts input / output requests issued to the general-purpose OS 5 by the processes 21 and 22 through the input / output acceptance interface 51. FIG. 2A is a conceptual diagram showing a format of an input / output request issued to the general-purpose OS 5. The input / output request issued to the general-purpose OS 5 includes a header 61 and input / output block information 62. The header 61 includes a process identifier 63 for identifying the issuing source process, and issuing source process information 64 indicating the class of the issuing source process. In the present embodiment, two classes of a real time class and a time share class are allowed for the process executed by the application 2. The issuer process information 64 describes whether the process that issued the input / output request is a real time class or a time share class. Here, the real-time class is a class that should be processed in real time, and the time share class is a class that does not require real-time processing. The input / output request issued to the general-purpose OS 5 is delivered to the real-time OS 3 by the inter-OS communication driver 4. The input / output block information 62 is information for specifying an input / output block for which input / output is requested. For example, an input / output block address is described.

  The real-time OS 3 includes an input / output reception interface 31, and receives input / output requests issued to the real-time OS 3 by the tasks 11 and 12 through the input / output reception interface 31. FIG. 2B is a conceptual diagram showing a format of an input / output request issued to the real-time OS 3. The input / output request issued to the real-time OS 3 includes a header 71 and input / output block information 72. The header 61 includes a task identifier 73 for identifying the issuer task, priority information indicating the priority of the issuer task, and issuer task information 75 indicating the class of the issuer task. Similar to the process executed by the application 2, two classes of the real time class and the time share class are allowed for the task executed by the application 1. The issuer task information 75 describes whether the issuer task is a real time class or a time share class. The input / output block information 72 is information for specifying an input / output block for which input / output is requested, and describes, for example, the address of the input / output block.

  Input / output processing for the shared disk 6 is performed by the priority controller 32 and the input / output processing interface 36.

  The priority controller 32 has the following three functions: First, the priority controller 32 has a function of dividing an input / output request issued to the general-purpose OS 5 into a plurality of input / output requests. Have. Specifically, when a certain input / output request issued to the general-purpose OS 5 requests input / output of a certain input / output block, the priority controller 32 divides the input / output block into a plurality of partial blocks. Multiple input / output requests that request to input / output.

  Second, when the priority controller 32 divides the input / output request issued to the general-purpose OS 5 into a plurality of input / output requests, the priority controller 32 sets the size of the partial block to the task that issued the input / output request to the real-time OS 3. It has a function to adjust in response to the priority. In adjusting the size of the partial block, the priority management table 33 and the partial block size designation table 34 are referred to. Here, as shown in FIG. 3A, the priority management table 33 is a table in which the task identifier of a task that has issued an input / output request to the real-time OS 3 and the priority of the task are registered. is there. It should be noted that the priority registered in the priority management table 33 is the priority described in the priority information 74 of the input / output request issued to the real-time OS 3. On the other hand, the partial block size designation table 34 is a table showing the correspondence between the priority of the task that issued the input / output request to the real-time OS 3 and the partial block size, as shown in FIG. 3B. .

  Thirdly, the priority controller 32 has a function of enqueuing each of the input / output requests to any of the input / output request queues # 1 to # 3 according to the nature of the issuing task or process. The input / output request queue # 1 is a queue given the highest priority, and is associated with the priority group # 1. Similarly, the input / output request queue # 2 is a queue to which the next highest priority is given, and is associated with the priority group # 2. The input / output request queue # 3 is a queue given the lowest priority, and is associated with the priority group # 3.

  The input / output request queue to be enqueued is selected with reference to the priority group management table 35. As shown in FIG. 4, the priority group management table 35 describes the nature of the issuer task or process belonging to each priority group. In the present embodiment, among the tasks executed on the real-time OS 3, those belonging to the real-time class are classified into the priority group # 1, and among the processes executed on the general-purpose OS 5, those belonging to the real-time class are It is classified into priority group # 2. Of the tasks executed on the real-time OS 3 and the processes executed on the general-purpose OS 5, the tasks or processes belonging to the time share class are executed on either the real-time OS 3 or the general-purpose OS 5, Grouped into priority group # 3. A certain input / output request is enqueued in the input / output request queue # 1 when the issuing task or process belongs to the priority group # 1. When the issuing task or process belongs to the priority group # 2, the input / output request is enqueued to the input / output request queue # 2, and when the issue task or process belongs to the priority group # 3, the input / output request is It is enqueued in the I / O request queue # 3.

  The input / output processing unit 38 performs input / output processing to the shared disk 6 in response to the input / output requests enqueued in the input / output request queues # 1 to # 3. The input / output processing is performed in order from the input / output request queue to which high priority is given. That is, the input / output process for the input / output request enqueued in the input / output request queue # 1 is performed first, and the input / output process for the input / output request enqueued in the input / output request queue # 2 is performed next. Next, the input / output processing for the input / output request enqueued in the input / output request queue # 3 is performed. Among the input / output requests enqueued in the same input / output request queue, the input / output processing is performed in order from the input / output request previously enqueued in the input / output request queue. When the input / output processing in response to each input / output request is completed, the input / output processing unit 38 notifies the task or process that issued the input / output request of the completion of the input / output processing.

  Next, the operation of the embedded system 10 in this embodiment will be described. FIG. 5 is a flowchart showing a procedure from issuing an input / output request to enqueuing the input / output request in the operation of the embedded system 10 in the present embodiment.

  When an input / output request is issued from the tasks 11 and 12 of the application 1 or the processes 21 and 22 of the application 2, the issued input / output request is sent to the input / output reception interface 31 of the real-time OS 3 or the input / output reception interface 51 of the general-purpose OS5. (Step A1). When an input / output request is received by the input / output reception interface 51, the input / output request is delivered to the real-time OS 3 by inter-OS communication. The input / output request received in step A1 is hereinafter referred to as a target input / output request.

  The priority controller 32 confirms whether the target input / output request is an input / output request issued to the real-time OS 3 or an input / output request issued to the general-purpose OS 5 (step A2). When the target input / output request is an input / output request issued to the real-time OS 3, the priority controller 32 refers to the task identifier 73 and the priority information 74 from the header 71 of the input / output request, and The task and the priority of the task are registered in the priority management table 33 (step A8). After the task and priority are registered, the process proceeds to step A7. When the target input / output request is an input / output request issued to the real-time OS 3, the target input / output request is not divided.

  On the other hand, when the target input / output request is an input / output request issued to the general-purpose OS 5, the priority controller 32 refers to the priority management table 33 (step A3) and sets the highest registered priority. Obtain (step A4). In other words, the priority controller 32 acquires the priority of the task having the highest priority among the tasks registered in the priority management table 33. Further, the priority controller 32 refers to the partial block size designation table 34, acquires the partial block size corresponding to the highest priority (step A5), and uses the acquired partial block size to execute the target input / output request. Is divided (step A6). By dividing the input / output request issued to the general-purpose OS 5 into a plurality of input / output requests, the execution time of the input / output processing per one input / output request is shortened, and the start of the input / output processing of the real-time OS 3 is delayed. Can be reduced.

  Specifically, when the priority controller 32 requests input / output of an input / output block having a target input / output request, the priority controller 32 designates the input / output block as a plurality of partial blocks having the partial block size acquired in step A5. A plurality of input / output requests for requesting input / output in a divided manner are generated. The contents of the input / output request issued to the general-purpose OS 5 are inherited by a plurality of input / output requests generated by the division in step A6, except for the contents of the input / output block information 62. That is, the plurality of input / output requests generated by the division in step A6 have the same contents as the input / output requests issued to the general-purpose OS 5 shown in FIG. 2A except for the contents of the input / output block information 62. is doing. On the other hand, the contents of the input / output block information 62 are rewritten to information for specifying each partial block.

  The partial block size used for dividing the target input / output request is determined so as to increase as the priority of the task having the highest priority among the tasks registered in the priority management table 33 increases. According to such control, since the partial block size is increased when there is no task that requires a high-speed response in the real-time OS 3, it is possible to suppress a decrease in throughput of a process executed on the general-purpose OS 5. .

  After dividing the target input / output request, the process proceeds to step A7.

  In step A7, the priority controller 32 enqueues the target input / output request or the input / output request generated by dividing the target input / output request into any of the input / output request queues # 1 to # 3. The input / output request queue to be enqueued is selected by comparing the issuer process information 64 or the issuer task information 75 described in the header 61 or 71 of the input / output request with the priority group management table 35. Each input / output request (target input / output request or input / output request generated by dividing the target input / output request) is issued when the issuer belongs to priority group # 1, that is, the issuer is on real-time OS 3 If the task is to be executed and the task belongs to the real-time class, it is enqueued in the input / output request queue # 1. Further, when the issuer of the input / output request belongs to the priority group # 2, that is, when the issuer is a process executed on the general-purpose OS 5 and the process belongs to the real-time class, the input / output request is Are enqueued in the input / output request queue # 2. When the issuer of the input / output request belongs to the priority group # 3, that is, when the issuer task or process belongs to the real-time class, the issuer task or process is on either the real-time OS 3 or the general-purpose OS 5. Regardless of whether it is executed, the input / output request is enqueued in the input / output request queue # 3. The input / output requests enqueued in the input / output request queues # 1 to # 3 are sequentially dequeued, and input / output processing for the shared disk 6 is performed in response to the dequeued input / output requests.

  The priority control is performed according to the nature of the issuer of the input / output request, that is, the input / output request contention control for the embedded system 10 as a whole is performed by selecting the input / output request queue to which the input / output request is to be enqueued. Is done optimally.

  FIG. 6 is a flowchart showing a procedure for performing input / output processing to the shared disk 6 in response to input / output requests enqueued in the input / output request queues # 1 to # 3. The input / output processing interface 36 checks whether the input / output request queue # 1 is empty (step B1). If it is not empty, the input / output processing interface 36 dequeues the first input / output request (the first input / output request enqueued) from the input / output request queue # 1 (step B4), and responds to the dequeued input / output request. The input / output process is performed (step B7).

  If the input / output request queue # 1 is empty, the input / output processing interface 36 checks whether the input / output request queue # 2 is empty (step B2). If it is not empty, the input / output processing interface 36 dequeues the first input / output request (the first input / output request enqueued) from the input / output request queue # 2 (step B5), and responds to the dequeued input / output request. The input / output process is performed (step B7).

  When both the input / output request queues # 1 and # 2 are empty, the input / output processing interface 36 checks whether or not the input / output request queue # 3 is empty (step B3). If not empty, the input / output processing interface 36 dequeues the first input / output request (the first input / output request enqueued) from the input / output request queue # 3 (step B6), and responds to the dequeued input / output request. The input / output process is performed (step B7).

  When the input / output processing is completed, the completion of the input / output processing is notified to the issuer of the input / output request (task executed on the real-time OS 3 or process executed on the general-purpose OS 5) (step B8). In addition, when the issuer of the input / output request for which the input / output processing has been completed is a task executed on the real-time OS 3, the input / output processing interface 36 notifies the priority controller 32 to that effect. The priority controller 32 deletes the task and its priority from the priority management table 33 (step B9). With the above procedure, the input / output processing to the shared disk 6 is completed.

  According to the embedded system 10 of the present embodiment, the input / output request issued to the general-purpose OS 5 is divided into a plurality of input / output requests, thereby shortening the execution time of input / output processing per one input / output request. Become. Therefore, the embedded system 10 of the present embodiment can reduce the delay in starting the input / output processing of the real-time OS 3. Further, in the embedded system 10 of this embodiment, the input / output request issued by the task executed by the real-time OS 3 is not preferentially processed unconditionally, but is given priority according to the nature of the input / output request issuer. Degree control is performed. Thereby, the competition control of the input / output processing as the entire embedded system 10 is optimally performed. In addition, in the embedded system 10 of this embodiment, the partial block size used for dividing the I / O request is adjusted according to the priority of the task executed on the real-time OS 3. According to such control, when there is no task that requires a high-speed response in the real-time OS 3, the partial block size can be increased, and a decrease in throughput of a process executed on the general-purpose OS 5 can be suppressed.

FIG. 1 is a block diagram showing a configuration of an embedded system according to an embodiment of the present invention. FIG. 2A is a conceptual diagram showing a format of an input / output request issued by a process executed by the general-purpose OS. FIG. 2B is a conceptual diagram showing a format of an input / output request issued by a process executed by the general-purpose OS. FIG. 3A is a conceptual diagram showing the contents of the priority management table. FIG. 3B is a conceptual diagram showing the contents of the partial block size designation table. FIG. 4 is a conceptual diagram showing the contents of the priority group management table. FIG. 5 is a flowchart showing a procedure from issuing an input / output request to enqueuing the input / output request in the present embodiment. FIG. 6 is a flowchart showing a procedure for performing input / output processing to the shared disk in response to an input / output request enqueued in the input / output request queue.

Explanation of symbols

10: Embedded system 1, 2: Application 3: Real-time OS
4: Inter-OS communication driver 5: General-purpose OS
6: Shared disk 11, 12: Task 21, 22: Process 31: Input / output acceptance interface 32: Priority controller 33: Priority management table 34: Partial block size designation table 35: Priority group management table 36: Input / output processing Interface 51: Input / output acceptance interface 61: Header 62: Input / output block information 63: Process identifier 64: Issuer process information 71: Header 72: Input / output block information 73: Task identifier 74: Priority information 75: Issuer task information

Claims (8)

A computer system in which a hybrid OS consisting of a real-time OS and a general-purpose OS is installed,
A shared device shared by the real-time OS and the general-purpose OS;
A plurality of requests for inputting / outputting a first input / output request issued to the general-purpose OS by dividing the input / output block requested for input / output by the first input / output request into a plurality of partial blocks. An input / output request dividing means for dividing the second input / output request;
Input / output processing means for performing input / output processing to the shared device in response to the plurality of second input / output requests and a third input / output request issued to the real-time OS;
The input / output processing means is configured to input the input to the shared device according to a property of an issuer of the first input / output request and an issuer of the third input / output request corresponding to the plurality of second input / output requests. A computer system that controls the priority of output processing. The computer system according to claim 1,
The plurality of second input / output requests and the third input / output request are made into the nature of the issuer of the first input / output request and the issuer of the third input / output request corresponding to the plurality of second input / output requests. Further comprising priority control means for enqueuing to any of a plurality of input / output request queues having different priorities according to
The input / output processing means enqueues an input / output process in response to an input / output request enqueued in the input / output request queue given a high priority into the input / output request queue given a low priority. A computer system that has priority over I / O processing in response to I / O requests. A computer system according to claim 2,
The third input / output request whose issuer is a real-time class is enqueued in the first input / output request queue having the highest priority among the plurality of input / output request queues,
The plurality of second input / output requests whose corresponding issuers of the first input / output requests are a real-time class are grouped into a second input / output request queue having a lower priority than the first input / output request queue. ,
The third input / output request in which the issuer is a time share class, and the plurality of second input / output requests in which the issuer of the corresponding first input / output request is a time share class are the second input / output requests A computer system that is grouped into a third I / O request queue that has a lower priority than the request queue. A computer system according to any one of claims 1 to 3,
The input / output request dividing means adjusts the size of the plurality of partial blocks according to the priority of a task that has issued the third input / output request to the real-time OS. A computer system according to claim 4,
Furthermore,
When the third input / output request is issued to the real-time OS, registration means for registering the task that issued the third input / output request and its priority in a table;
The task that issued the third input / output request for which the input / output processing for the shared device has been performed, and deletion means for deleting the priority from the table;
The size of the plurality of partial blocks is adjusted according to a priority of a highest priority task which is a task having the highest priority among the tasks registered in the table. A computer system according to claim 5,
The size of the plurality of partial blocks is adjusted such that the higher the priority of the highest priority task, the smaller the size of the plurality of partial blocks. An input / output priority control method in a computer system in which a hybrid OS comprising a real-time OS and a general-purpose OS is installed and a shared device is shared by the real-time OS and the general-purpose OS,
A plurality of requests for inputting / outputting a first input / output request issued to the general-purpose OS by dividing the input / output block requested for input / output by the first input / output request into a plurality of partial blocks. Dividing into second I / O requests;
Performing an input / output process to the shared device in response to the plurality of second input / output requests and a third input / output request issued to the real-time OS,
The step of performing the input / output processing to the shared device is performed according to the nature of the issue source of the first input / output request and the issue source of the third input / output request corresponding to the plurality of second input / output requests. An input / output priority control method comprising a step of performing priority control of the input / output processing. In a computer system in which a hybrid OS composed of a real-time OS and a general-purpose OS is installed and a shared device is shared by the real-time OS and the general-purpose OS,
A plurality of requests for inputting / outputting a first input / output request issued to the general-purpose OS by dividing the input / output block requested for input / output by the first input / output request into a plurality of partial blocks. Dividing into second I / O requests;
In response to the plurality of second input / output requests and a third input / output request issued to the real-time OS, a step of performing input / output processing to the shared device and a program to be executed. In the step of performing the input / output processing, priority control of the input / output processing to the shared device is performed by issuing the first input / output request corresponding to the plurality of second input / output requests and the third input / output. A program that depends on the nature of the issuer of the request.

Images