JPH02297620A - Input/output control system - Google Patents

Abstract

PURPOSE:To prevent data change according to mis-positioning in a rerun operation from occurring by comparing a designated record passing finally with the final record passing completely on the recording path of a write/readout mechanism. CONSTITUTION:Record detecting means 41 provided at direct access memory devices i.e. magnetic disk devices 31-34 detect the records passing completely on the recording paths of the write/readout mechanisms of the magnetic disk devices. Also, comparison means 42 provided at the magnetic disk controllers 21 and 22 compare the designated records passing finally with the records passing finally detected by the record detecting means 41 with a command, and confirm whether or not they coincide with each other. Thereby, it is possible to prevent the data change according to the mis-positioning in the rerun operation from occurring.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application This invention relates to a human output control method for controlling the transfer of data between a central processing unit (CPU) and a direct access storage device via channels and control devices. In particular, after detecting a retryable error on a circulating recording path (track) in the above-mentioned direct access storage device, it is possible to reliably position the record to be retried and to indicate that it has been written or read by a command. This is related to the input/output control method to be confirmed.

[Prior Art] FIG. 3 is a diagram schematically showing a conventional input/output control system. 11 is a central processing unit (CPU), 12 is a channel processing device, 21.22 is a magnetic disk control device, 31 to 3
4 is a magnetic disk device. 13 in the channel processing device 12 is a channel.

In the input/output control system as shown in FIG.
3 and the magnetic disk control device 21 or magnetic disk device 33.3 that controls the magnetic disk device 31.32.
4 is connected via a magnetic disk control device 22 that controls the 4.

The magnetic disk control device 21 temporarily stores and transfers data according to a channel command word (hereinafter referred to as CCW) sent from the central processing unit 11 via the channel 13, and also controls the magnetic disk devices 31 to 34. give commands to. This command is the magnetic disk device 31-3
This is to control data storage and transfer in 4 and perform desired operations.

In the magnetic disk device used in such a human output control system, the array format (format) stored on each track of the disk serving as the recording medium is the count key data format ( CKD below
(format) is common. In the CKD format, there is an index marker a at the beginning of each track, followed immediately by a home address (HA) and record O (RO).

These are mandatory regardless of the user.

Cylinder numbers, track numbers, etc. are recorded in the HA.

The RO is managed by the system program and consists of a count section and a data section. Data section R immediately after RO
There are 1, R2, R3, . . . a, and Rn, which confirm user data and consist of a count section, a key section, and a data section. Index markers a and HA, HA and RO, and each record are separated by gaps. The last a is the same as the index marker a and is also the beginning of one rotation.

[Problems to be Solved by the Invention] In such a magnetic disk device, errors such as data overrun may occur in the magnetic disk controller 21. during data transfer to rewrite a desired data portion Ri.
22 is detected, the magnetic disk controller 21.22
does not immediately report the error to the program by generating a human output interrupt to the central processing unit 11 via the channel 13, but performs a retry with the channel 13 as much as possible. In this way, the efficiency of the entire system is improved by not reporting to the program.

Since the retry is the same CCW re-execution from the beginning of the data section Ri where the error was detected, it takes approximately one rotation from the time of error detection to the start of the retry after positioning the write/read mechanism to the data section Ri. A little waiting time (several tens of meters)
ec) is required. If the connection with channel 13 is continued during this period, the connection with channel 13 is normally disconnected once an error is detected, in order to avoid reducing channel utilization efficiency due to loss of connection opportunities for other magnetic disk devices on channel 13. Immediately before the data portion Ri, a recombination request is made to the channel 13 and a retry is performed.

Therefore, the magnetic disk controllers 21 and 22 are provided with a mechanism for determining when to recombine.

However, bugs related to the above mechanism cannot be eliminated in reality.

The reason for this is that since errors occur accidentally, it is impossible to conduct a simulated failure test with comprehensive timing etc. set in advance. However, although the absolute number of such bugs is small, once they occur, they may cause data corruption, so they have a large impact on the system. moreover,
Even if a program is able to detect data corruption by chance, it often takes a long time to find the cause because it is difficult to reproduce the error that caused the bug.

Regarding the storage of database information in the magnetic disk devices 31 to 34, it is common for the data length of each record on a track to be constant within a logical file. That is, in files of the same type, the data length of each data portion R1 to Rn is constant, for example, 4096 bytes. When backing up the database information stored in the magnetic disk devices 31 to 34 to a magnetic tape device, etc., once the database information is read from the magnetic disk devices 31 to 34 to the I10 buffer area in the main storage device, the database information stored in the I10 buffer area is Data is written to a magnetic tape device, etc., and the general CCW sequence used for controlling the former at this time is shown in Figure 5 (a).
Shown below. In order to increase speed, the CCW column is not read in units of records, but multiple records on one track are read out all at once by one input/output command. In this CCW column, 12 CCWs from ■ to @ are command chained. In other words, the first 5EEK command sets the write/read mechanism to the track with the desired cylinder number and head number, the second SET 5ECTOR command positions the write/read mechanism just before RO, and the third 5EAR
Use the CHID EQUAL command to check if it is the count part of RO, and if it is, the 5th RE
Specify the data part of RO with the AD DATA command.
Read into 10 buffer areas, and then read the 6th READ
The count part, key part, and data part of the next record on the track are read into the 10 buffer area with the CKD command, and the same operation is repeated with the 7th to 12th READ CKD commands.

FIG. 5(b) shows the result of storing data in the I/○ buffer area when the CCW sequence is executed normally.

An error occurred while the 8th READCKD command in the CCW column in FIG. The data section R2 is in the buffer area.
is read out twice, and the data section R7 is removed. Figure 5 (C) shows the result of data storage in the I10 buffer area in this case.
Shown below. However, the execution result of the CCW is reported to the program as normal. Unless the program checks the contents of the I10 buffer area, this data corruption will not be detected, and as a result, it will be written as is to the magnetic tape device. Even if the program fortunately detects a logical mismatch when the magnetic tape is read after a period of time, it will take a considerable amount of time to find the cause, as various bug factors are assumed. If a program is unable to detect a logical inconsistency when reading the magnetic tape, the data corruption may have an even wider adverse effect.

In a direct access storage device such as a magnetic disk device, after detecting a retryable error in a circulating recording path (track), the present invention reliably positions the record to be retried and completes writing or reading. By checking what you have done with the command,
The object of the present invention is to provide an input/output control method that can prevent data corruption as described above due to positioning errors at the time of retry.

[Means to solve the problem]

In the input/output control method according to the present invention, the direct access storage device has a record detection means for detecting a record that has passed completely on the recording path of the write/read mechanism, and further detects a record that has passed the last time specified by a command. It holds a comparison means for comparing it with the last record that has passed completely on the recording path of the writing/reading mechanism detected by the means to confirm whether or not they are the same.

[Effect]

In the present invention, the direct access storage device detects the last completely passed record on the record path of the write/read mechanism, and operates normally when this last record is the same as the last passed record specified by the command. It is confirmed that

〔Example〕

FIG. 1 is a block diagram showing one embodiment of the inventor's output control method. In this figure, 41 is a record detection means,
It is provided in direct access storage devices, that is, magnetic disk devices 31 to 34. The record detecting means 41 detects a record that has completely passed on the recording path of the write/read mechanism of the magnetic disk device. Also, 4
Reference numeral 2 denotes comparison means, which is provided in the magnetic disk control devices 21 and 22. This comparison means 42 compares the record of the last passage specified by the command with the record of the last passage detected by the record detecting means 41, and confirms whether the two are the same.

Other mechanisms are the same as the conventional example shown in FIG. Note that the comparing means 42 may be provided at a location other than the magnetic disk controllers 21 and 22.

Next, the operation will be explained.

Immediately after positioning on the recording path (track) designated by the writing/reading mechanism of the magnetic disk devices 31 to 34, the record detecting means 41 detects and detects the record number on which the writing/reading mechanism has completely passed on the track. Remember.

For convenience, the record number stored by the record detecting means 41 will be referred to as the "last passed record number." moreover,
A command for checking the last passed record number is added as a command function for the magnetic disk devices 31 to 34. For convenience, use this command as rcHEcK LAS
T RECORD NUMBER (CHECK LR
) Call it J.

The CHECK LR command sends the cylinder number, head number, and record number data prepared by the program to the magnetic disk controller 21.22, and checks the specified magnetic disk devices 31 to 34 in the magnetic disk controller 21.22. A comparison means 42 compares the cylinder number, head number, and last passed record number currently positioned in the write/read mechanism with the output data. The magnetic disk controllers 21 and 22 that have received the CHECK LR command respond with normal termination (channel end and device end) as a device status byte (DSB) if the comparison results are the same. If the comparison results are not the same, the DSB responds with abnormal termination (device check, channel termination, and device termination), and prepares detailed information on the abnormality as a sense byte.

An example of using the record detecting means 41 and comparing means 42 to notify the program of the data corruption shown in FIG. 5(C) at the time of its occurrence will be described below.

Figure 2 shows the CHECK for the CCW column in Figure 5(a).
The application of the LR command is shown. CCW in Figure 5(a)
Command chain the CHECK LR command for the column at the end. The data prepared by this command has the same cylinder number, head number, and record number "7" as those prepared by the first 5EEK command. In the CCW column of FIG. 2, even if a malfunction like the one shown in FIG. is "7", whereas the last passed record number is "6", so a mismatch is detected and the DSB is responded to as abnormal termination. As a result, the program knows that the input/output operation shown in FIG. 2 is abnormal, and can avoid data corruption by performing appropriate recovery processing such as retrying.

Note that although the above example describes the time of reading from the magnetic disk devices 31 to 34, the same applies to the time of writing. Furthermore, although the magnetic disk control devices 21 and 22 are for controlling the magnetic disk devices 31 to 34, they may be used as control devices when the magnetic disk devices 31 to 34 are other direct access storage devices.

〔Effect of the invention〕

As explained above, the present invention provides an input/output control system that includes a record detection means for detecting the last record that the direct access storage device has completely passed on the recording path of the write/read mechanism, and further provides a The last passed record specified by is compared with the last record passed completely on the recording path of the write/read mechanism, and a comparison means is maintained to check whether they are the same. This has the advantage that garbled data due to positioning errors can be reliably prevented.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.
FIG. 2 is a diagram showing an example of a CCW train according to the present invention, FIG. 3 is a diagram showing an outline of a conventional human output control subsystem,
Figure 4 is a diagram showing the data arrangement format (CKD format) on one track of a magnetic disk unit, and Figure 5 (a) shows how multiple records on one track of a magnetic disk unit are collectively processed by one input/output command. Read to the I10 buffer area in main memory. FIG. 5(b) is a diagram showing the CCW column generally used when reading out data, and FIG. Figure (e) is a diagram showing the result of data storage in the I10 buffer area in the case of an abnormality as well. In the figure, 11 is a central processing unit, 12 is a channel processing unit,
13 is a channel, 21.22 is a magnetic disk controller,
31 to 34 are magnetic disk devices, 41 is a record detection means, and 42 is a comparison means. Figure 2 Figure 5 (a)

Claims (1)

[Claims] a channel and a direct access storage device, and a controller interconnecting the channel and the direct access storage device, the channel connecting the direct access storage device for sequentially providing a series of channel commands to the controller. The control device and the direct access storage device are controlled to issue the command to the direct access storage device after executing the command to the channel, and the control device and the direct access storage device are configured to notify the channel of an end signal after each execution of the command. controlled, the direct access storage device is capable of recording a plurality of records each consisting of a plurality of independent fields along a circular recording path of a relatively movable medium, and positioning the write/read mechanism for the records. If the direct access storage device detects an abnormality after receiving the command and before reporting an end signal of the operation to the specified record to the channel, the direct access storage device In the input/output control system, the direct access storage device is controlled to enable re-execution of the same command with the device by positioning the write/read mechanism at the beginning of the same record. record detection means for detecting the record that has passed completely on the recording path of the mechanism; An input/output control method characterized by comprising a comparison means for comparing the record with the last record completely passed through to confirm whether or not the record is the same.