JPS62274448A - Control system for input/output device - Google Patents

Abstract

PURPOSE:To share a file without reducing the system throughput by making it possible for each OS to change and test control state information as a part of the input/output operation and detecting the change of control state information without testing control state information by the OS. CONSTITUTION:During the copying operation, the whole of disc devices 131 and 142 is not closed but only current copying areas 413 and 414 of respective disc devices and an uncopied area 416 of a disc device 142 are closed. Consequently, normal input/output operations other than the copying operation are executed by dual operation while the copy operation to copied areas 411 and 412 of both disc devices is performed, and the identify of data contents is kept. Execution of normal input/output operations other than the copying operation to copying areas 413 and 414 is suspended until the copying operation to their cylinders is terminated. Normal input/output operations other than the copying operation to the uncopied area 415 are executed by single operation of only the disc device 131.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a control method for an input/output device used in an information processing system when the input/output device is shared by a plurality of operating systems.

[Conventional technology]

Information processing systems are sometimes configured using multiple operating systems (hereinafter simply referred to as OS) in order to improve the reliability and processing capacity of the system. At this time, the input/output device is jointly used by a plurality of OSs due to the need for data processed by one OS to be processed by another OS.

In such a case, there are 10! management states of the input/output devices. I
Even if a change is made from the 08VC of the 08VC, the change is not immediately transmitted to all other O8I/C@s, so the management state information of the input/output devices stored in the central processing unit of each O8K may not match (become .For this reason, after changing the management status of the input/output device K, another O8 that has the management status information before the change,
When an input/output command is executed to the input/output device RIIC whose management state has been changed, there is a problem in that the appropriate input/output operation cannot be executed.

In addition, in order to change the management state of a certain function of a shared input/output device, while an O8 is making complex preparations such as initial settings for that input/output device, in other words, this preparation O3 is making preparations until the
It is necessary to prevent other O8s from using this function.

In order to avoid the above-mentioned problems and satisfy the above-mentioned needs, when changing the management status of input/output devices, there is a method of providing an inter-channel coupling device or the like to communicate between O8s. . However, this method has the disadvantage that there is a delay due to the time required for communication, and that it takes time to change the management state information of the input/output devices in each O8. Further, if the O8 to be contacted is down, special consideration is required because appropriate contact cannot be made.

Instead of communicating between the O8s mentioned above, the physical status information is stored in an external storage device such as a disk device that is shared by multiple O8s, and the K'f state information is stored when inputting/outputting to the input/output device.
There is also a method in which the management status information stored in the external storage device is always tested, and the input/output operation is executed only when there is no change in the information. However, this method is not practical because it requires time to successively retrieve and test the management status information from the external storage device each time a normal input/output operation is performed on the input/output device.

Therefore, a method may be considered in which the management state information of the input/output device is stored in the input/output device or the input/output control device lK. This type of prior art is described in Japanese Patent Laid-Open No. 180765/1983. The conventional technology disclosed in this publication is capable of quickly making the contents of both disk devices identical by installing a memory in the input/output control device when the contents of the disk devices serving as the redundant input/output device are no longer the same. It is something that becomes.

However, in this prior art, no consideration is given to the case where a plurality of O8s jointly use an input/output device.

[Problem that the invention seeks to solve]

As mentioned above, in the conventional technology, in an information processing system using multiple O8s, input/output operations from the O8s may not be able to be performed properly, or input/output operations take a long time. There were drawbacks such as this.

An object of the present invention is to provide an input/output device that enables the management status of an input/output device shared by multiple O8s to be detected without reducing system throughput without requiring multiple O8s to communicate directly with each other. Obtaining a control method K
be.

[Means for solving problems]

According to the present invention, the purpose is to store the management state of the input/output device shared by the plurality of O8s as management state information in the input/output device or the input/output control device shared by the plurality of O8s. Means are provided to enable each O8 to modify and test this administrative state information as part of its I/O operations, and further means for communicating the changed administrative state information to the O8s to which the I/O device is connected. K is achieved by providing an input/output device in the input/output device so that the O8 can detect a change in the management state information without testing the management state information in the input/output device or the input/output control device.

[For production]

In the present invention, the storage means provided in the input/output device [or the input/output control device] stores management state information regarding the corresponding input/output device. By testing this storage means each time each O8 performs an input/output operation,
Since it is possible to detect changes in the management status of the input/output device and update its own management status information, other O8
Even if the management status of the input/output device has been changed by
Processing can be executed in response to changes in management status.

〔Example〕

Hereinafter, a control method for an input/output device according to the present invention will be explained in detail with reference to illustrated embodiments, using a disk device as a specific example of the input/output device.

FIG. 1 is a block diagram showing an embodiment of the present invention when a plurality of O8s share a disk device, and FIG. A diagram showing an example of a memory map of a memory portion corresponding to one disk device ffl, FIG. 3 is a diagram showing a command chain, and FIG. 4 is a diagram showing a case where the contents of duplicated disk devices become inconsistent. FIG. 3 is a diagram showing a cylinder map for explaining a copy operation for matching the contents.

In Figure 1, 03101 is channel 111°112
O8102 Hachi? It is connected to disk controllers 121 and 122 via channels 113 and 114, and the disk controllers 121 and 122 are connected to a plurality of disk drives [131, 132 and 141, 142, respectively. Further, the disk control devices R121 and 122 are equipped with memories 151 and 152 that store the management status of each of the disk devices they control as -W embedding information.

This W filling status information is information necessary for the O8 to manage each disk device, and is normally stored in the O8, but a copy of this information is stored in the memories 151 and 152.
It will also exist within. The memory 151 includes a disk controller [all disk devices 131, 132 under 121].
The management state information for the memo '7152K is for all disk devices 141, 142 under the disk control device 122.
Management state information for each is stored and stored. The memories 151 and 152 are preferably ones that can be accessed at high speed, and semiconductor memories are generally suitable. Further, the memories 151 and 152 are connected to the disk controller [121, 152].
122, but each disk drive [131, 132, 1
41 and 142, respectively.

A case in which the 08101 performs input/output operations to the disk device 131 will be described below.

08101 is channel 111. Disk control device 12
1 to access memory 151, in this case,
Memory 15 using the command chain shown in Figure 3.
Access 1. CHECK TABLE301 and READ TA are new commands for this invention.
CHECK TAB BLE 305 is available
The LE 301 determines whether the management status information in the memory 151 or 152 satisfies the conditions expected by the O8, that is, the management status information held in its own O8 corresponding to the input/output device and the management status information in the memory 151 or 152. READ is a command that tests whether the management status information of
TABLE 305 is a command for reading management state information in memory 151 or 152.

In order for the 08101 to perform input/output operations to the disk device lt131, the O8101 first executes CHECK TABLE301 in the command chain described above, and stores the management status information corresponding to the disk device [131] owned by the O8101 and the memory 151. A check is executed to see if the management status information corresponding to the disk device 131 in If a match is found, the O8 command execution is
A jump is made to the user command chain 304 by a known jump destination change command TIC 303, and the operation is performed on this user command chain 304. This user command chain 304 is
This is a command chain that executes a commonly used input/output operation, and by executing this command, the O8101 completes a predetermined input/output operation for the disk device 131.
Execute TABLE301 and save data in memory 151 and own O8
If a discrepancy is detected in the management status tif information for the disk devices in the 03IOI command, the command REA
Jumped to DTABLE305 and executed this command 30
This is done for 5. O8101 uses this command 30
By executing step 5, the management state information of the disk group e131 is read from the memory 151, and the read management state information is analyzed.

Also, the disk device 131 read from the memory 151
The management status information of the disk device 131 in the 08IOI
The information is taken into the O8101 as management status information.

The above operation also applies when 08102 executes input/output operations for disk group #131.
2 executes an input/output operation to a disk device other than the disk device 131 in exactly the same manner. In controlling the disk device as the input/output device described above, the management state information may be changed by a software command from the O8, or as part of normal input/output operations. Further, the notification of the change in the management state information may be triggered by the input/output device itself, or may be triggered by a command from another software.

In this way, 08IOI, 102 is; Cloak CHEC
By executing K TABLE 301, a predetermined input/output operation for each disk is executed only when the management state information of each disk device in the memory 151, 152 satisfies the expected conditions, and if not, the management state It is possible to detect changes in information and take appropriate action.

As a result of the above, 03101 and 03101, 08101.1 are specially able to change the management status of commonly used disk devices.
02 can be detected without communicating with each other. And, as already mentioned, the button, memory 151,1
52 uses semiconductor memory that can be accessed at high speed, and O8 can detect changes in the management state of disk devices as part of normal input/output operations, so it is possible to prevent a decrease in system throughput. .

Next, a case will be described in which the present invention is applied to a duplex disk device.

General K. Duplicating input/output devices is a means to improve the reliability of the entire information processing system, and the duplicated input/output devices always hold the same information.

Now, in FIG. 1, it is assumed that the disk devices 131 and 142 are duplicated.

In this case, a write operation from 08IOI or 102 is performed by 08101 or 102.
08101 or 102 is the disk group #131,
142, and the input/output command is issued and executed.

Such a dual disk family jilt 31゜1
When using 42 from a plurality of oSlols and lO2, the following problem occurs. That is, when 08101 writes @tatw data to a disk device and then attempts to write the same data to the disk device 142, if a failure occurs in the disk device 142,
Thereafter, at 08101°, only the disk family fl131 is used, and the disk drive 142 is logically blocked and no longer used, and the information processing system shown in FIG. 1 is effectively in a one-lung operation state. In this state, when 03102 performs a data read operation from disk group ftY:142 of one of the duplicated disk devices, 03102 has not yet detected the above-mentioned state of the disk device and reads data from disk group ftY:142. A problem arises in that old data that has not been written by 03101 is read and used.

According to the present invention, such problems can also be solved. The operation in this case will be described below with reference to FIGS. 2 and 4. In the memory map of management state information corresponding to one disk device shown in FIG. The copy status 7 lag 203 indicates whether the disk device is currently performing a copy operation to shift from single-lung operation to dual-line operation. Cylinder number 204 indicates the cylinder number in which a copy operation is currently being performed. Further, the input/output operation pending flag 205 indicates that the disk device has temporarily suspended the input/output operation since O8. Memo U 151 provided in the disk control device
, 152 can store management state information other than the above as necessary.

Now, suppose that the O8101 writes data to the disk device 131 and then writes the same data to the disk device 142.

As already mentioned, the 08101 executes the command chain shown in FIG. When adding data, data can be written to the disk device 131. Further, 08101 similarly reads the management status information corresponding to the disk group fFt142 from the memory 152 in the disk control device 12', and
An attempt is made to write the same data that was previously written to the disk device 131. At this time, if a failure occurs in the disk device 142, the 08101 cancels this write operation and changes the block state flag 202 of the management status information corresponding to the disk device 142 in the memory 152 from "off" to "
"On". As a result, the information processing system in FIG. 1 is changed from a double-lung operation state to a single-lung operation state, and one disk group #142 of the duplex disk device is changed.
is considered to be in a closed state.

After that, when the O8102 executes the command φ chain in FIG. 3 to read the data of the disk group R142, the O8102 expects that the blockage state flag 202 of the disk device 142 is "off", so the CH
The condition is not satisfied in ECK-TABLE 301, and user command chain 304 will not be executed. As a result, 08102 does not use old data that O3101 has not written, and also
It is possible to detect that the disk family fl142 has been blocked by OI and that the management state information has been changed, and as part of normal input/output operations, the changed management state information is imported into the own O8 and the management state information that it has is can be updated.

If the information processing system shown in FIG.
31 only, disk family ff113
The recorded contents of 1 and 142 are different. Therefore, in order to return the single-lung operation state to the double-lung operation state 11, K needs to read the contents of the disk device 131 to the disk device 142.

In the present invention, by using the memories 151 and 152, efficient copying operations can be performed and the disk device can be effectively blocked during the copying operation.

FIG. 4 shows the disk device 1 for explaining the copy operation.
31 and 142, each disk device 131 and 142 has nine cylinders (cylinder numbers 0-w3-1) indicated by areas 401 and 402, respectively. The copy operation is assumed to be executed in ascending order of cylinder numbers starting from cylinder number O, and assuming that copying is currently being executed from area 413 of cylinder number X to area 414, cylinder number 0-j-1 Area 41 of
1.412 has been copied and the disk device 131
The contents of this area of and 142 are the same. Areas 415 and 416 with cylinder numbers x+1 to rL-1 are last copy areas, and their contents may be different. This copy operation is executed by the OS 101 or 102, but if the copy is executed with the disk device 142 blocked, normal input/output operations other than the copy operation may cause the copied disk device 131 to be The contents of the area 411 may be rewritten.

In this case, the contents of areas 411 and 412 will be in different states, and it will not be possible to shift to the image frame operation state even if the copying operation is completed. In order to prevent this, writing operations other than normal input/output operations other than copying operations may be prohibited during copying, but this will result in a reduction in system throughput.

According to the present invention, during the copy operation, the disk group [131
, 142, but only the areas 413, 414 currently being copied in each disk device and the last copying area 416 in the disk device 142 are closed. As a result, the copied areas 411 and 412 of both disk devices
Normal input/output operations other than the copy operation during the copy operation for 1C can be executed by driving with both knees, and the identity of the data content is maintained. Area being copied 413.
Normal input/output operations other than copy operations for 414 are as follows:
Normal input/output operations other than the copying operation for the uncopied area 415, whose execution is suspended until the copying of that cylinder is completed, can be executed by operating only one side of the disk device 131.

08101 is the disk family [131 to the disk family [14
2, in case 08102 attempts to perform normal input/output operations other than the copy operation, 03101 uses memories 151 and 152 in disk controllers 121 and 122 to store the 131,1
Closing and opening/closing of 42 are executed as follows. 08101
is the management state information corresponding to the disk device 131 in the memory 151 and the disk group f! in the memory 152 prior to the copy operation. Among the management status information corresponding to t142, the copy status flag 203 is set to "on", and the cylinder number of the copying cylinder number 204 is set to "0". 0
8101 copies data with cylinder number "0" from disk device 131 to 142, and after copying is completed, increments the cylinder number of cylinder number 204 during copying by 1.
1" and continues the copy operation for this cylinder, and instructs disk controller #121 or 122 to release the blockage for cylinder number "0JYc.03
While 101 is copying data of cylinder number rOJ, 0
When 8102 initiates an input/output operation to cylinder number "0", the input/output operation pending flag 205 in the memory 151 in the disk control device 121 is set to "on", and 081
This I/O operation from 02 is suspended. When the disk controller 121 receives an instruction from 03101 to release the blockage for cylinder number rOJ, it checks the I/O operation pending flag 205 in the memory 151, and if it is "1 on", issues an I/O interrupt to 08102. reports that the blockage for cylinder number "0" has been released, and sets the I/O operation pending 7 lag 205 in the memory 151 to "
Turn off. By repeating the closing a of the cylinder during copying and the unblocking of the cylinder at the end of copying in this way, the entire area 40 of the disk group R131 is obtained.
1 is copied to the entire area 402 of the disk device 142. When all the copies are completed, the 08IOI turns off the copy status flag 203 of the management status information corresponding to the disk group ft131 in the memory 151, and turns off the copy status flag 203 of the management status information corresponding to the disk device 142 in the memory 152. Of these, copy status flag 203 and blockage status flag 2
Turn 02 "off". As a result, the information processing system shown in FIG. 1 is brought into a double-door operating state.

08101 sets the area 413 of cylinder number r-J to area 4
While the OS 102 is performing a copy operation to the disk device 131 . When attempting to perform normal input/output operations other than copy operations on 08102, v
f, the command chain shown in Figure 3 can be used. 08
102 indicates the disk device 131 [for which the copy status flag 203 is “on” and the copying cylinder number 204
However, in the expectation that the cylinder number is different from the cylinder number of the input/output operation that the OS 102 is about to execute, for the disk device 142, the copy status flag 203 is "on" and the copying cylinder number 204 is 08102. Execute CHECK TABLE 301, expecting the cylinder number to be greater than the cylinder number of the input/output operation to be executed. As a result, disk family f! At t131; input/output operations to the copying area 413 are not executed, but input/output operations to the copied area 411 and uncopied area 415Vc are executed, and input/output operations to the copying area 414 and uncopied area 416 of the disk device 142 are performed. No output operation is executed, and if an input/output operation is executed for the copied area 412, then K is reached.

During the above-mentioned copy operation from disk group f131 to disk group e142 by O8,101, 0S10
2 attempts to perform a normal input/output operation other than a copy operation on the disk device t131 or 142, but if the input/output operation is not executed and is suspended, the disk controller 121 or 122 The input/output operation pending flag 205 is set to "on". Since the cylinder number being copied changes as the copy operation progresses, the disk controller 121 or 122 sets the I/O operation pending flag 2 in the memory 151 or 152.
As already mentioned, when 05 is in the "on" trap, the unblocking of the disk device 131 or 142 is reported to 08102 using an input/output interrupt.

In the embodiment described above, the disk group [131 and 142
Although the explanation was made assuming that the disk drive 1 is duplicated,
The same applies when 32 and 141 are duplicated.

Further, in the copy operation between duplicated disk devices, the copy range is the entire disk device, but it is also possible to execute the copy operation only in a part of the area of the disk device. Furthermore, although the area of the disk device that is blocked during copying is assumed to be in units of cylinders, it may be made in units of magnetic heads of the disk device, that is, in units of tracks, by increasing the capacity of the memory provided in the disk control device.

〔Effect of the invention〕

According to the present invention, when the input/output bag y is shared by i number of O8s, a special communication means such as an inter-channel coupling device is provided between the O8s, and special communication between the O8s is established. It is possible to change the management status of an input/output device without any conflict, and to recognize changes in the management status of an input/output device by other 08VCs as part of normal input/output operations. For this reason, O
This has the great effect of significantly reducing the overhead required for communication between 8 and 8, and making it possible to share files without reducing system throughput or imposing any operational restrictions. Furthermore, according to the present invention, even in a copy operation when input/output devices are duplicated, it is possible to effectively block the input/output device during the copy operation without causing a decrease in system throughput. .

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention, FIG. 2 is a diagram showing an example of a memory map, FIG. 3 is a diagram showing a command chain, and FIG. F4 is a diagram showing a cylinder map. 101.102... Operating system, 111-114... Channel, 121,1
22...Disk control device, 131, 132,
141゜142...Disk device, 151,1
52...Memory. Figure 1 Figure 2i11J

Claims (1)

[Claims] 1. In an information processing system configured with a plurality of operating systems, an input/output device and an input/output control device shared by the plurality of operating systems, and an input/output device for inputting data to the input/output device. In order to store the management status of input/output devices including output as management status information,
The plurality of storage means provided in the input/output device or the input/output control device and the stored management state information are tested as part of the input/output operation prior to normal input/output operations for the input/output device. and a test means provided for each of the plurality of operating systems, so that each of the plurality of operating systems can perform processing according to the management state of the input/output device. Control method for input/output devices. 2. In the information processing system, the input/output device or the input/output control device is provided with means for notifying a connected operating system of a change in the management state of the input/output device. Range 1
Control method for input/output devices described in Section 1.