JPS5953946A - Error processing system in pushing-off control of store operation - Google Patents

Abstract

PURPOSE:To execute a complete restoration to an error generated in a pushing-off control of a store operation, by executing the processing to all contents of an FSA stack at the time of failing storage address FSA. CONSTITUTION:As for a write and read-out address to a buffer storage device BS, contents of an address register ADR, REG are read and held in a failing storage address FSA. In case when a new error is generated before the error processing to the address held in said stack is executed, an address to newly generated error is read in the following area of the FSA stack. Accordingly, the address corresponding to the error generated before the processing of FSA is all held in the FSA stack, and when the processing of FSA is executed, the contents of the FSA stack are all read out, and the processing to each of them is executed.

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to an error handling method during push-off control of a store operation in a storage device rt. Dragon: In a system with all storage devices, the main storage unit (M2O) and the storage and channel 1h control unit (8CU/C
Since the store operation between HCs is slow, it is common practice to increase the processing speed and performance of the system by using a push-off control sound for the store operation. , In the system of FIG. 1, a plurality of channels C)j, o, CI41...Cf-1x are
Storage and channels'ili! -(1 unit)
(SCU/CHC) and stores data from the channel to the main storage unit (SCU/CHC). and then stored from the buffer Strano unit (13s) to the main storage unit (MSU), if a store operation is performed, e.g. channel C) sent from io The coming data is MS
UK, channel Ct- at the appropriate timing (BS [when data is stored, etc.])
1o to signal the end of the store, and from I3S to Ail, where the data is stored in the MSU.

Perform the store end operation in SCU/CE-1c and store operation from channel C) J, 0 until it is changed.
There is no need to run il'+IGA, and the performance of the female system increases by IDI.

By the way, the channel and SC that sent the data like this
If an error is detected in the data stored in the BS from the channel after U is disconnected, the error cannot be reported to the corresponding channel and the data cannot be completely corrected. That is, after the channel store has been allocated, it is not possible to deal with errors when access to another channel has started. Also.

Even when there is no access from another channel, if the operation of the first channel has already ended normally,
Unable to report error occurrence to that channel. In order to deal with such an interval, the conventional store operation is separated from it',! In the control, the channel and S
If an error occurs regarding data stored from a channel in a remote channel when the CU is disconnected, the address of the data is FSK.
(Fa i 1 ing Storage Addre
ssl is held in the register, and when the error is accepted as an interrupt, read register f of this FSA and disconnect the corresponding address and continue processing unrelated parts. It is configured as follows.

However, in the foregoing control, if a similar error occurs before i'8A is read, for example, an error related to the data stored from the instruction processing unit (IPU), F8Ai Because the address written in the holding register is destroyed, it becomes impossible to know whether it was an FSA or not.Therefore, in this case, it is impossible to know which address caused the error.
There is a time when all operations are unavoidable, and as a result, the CPU of the system itself enters a check-stop state. This situation poses a serious problem, especially in online systems.

(3) Object of the Invention In view of the above-mentioned problems of the conventional method, an object of the present invention is to provide a stack of FSAs in the push-off control of the store operation, thereby eliminating all the problems that occur before the processing of 91"8A. Keep FSAi about errors and FS
8(4) Structure of the Invention It is an object of the present invention to provide an error handling method that enables processing of all contents of the F8A stack when processing A. Unit, storage and channel system? In the p4n'L control of store operation in a storage device with one hit unit and multiple channels,
Igj≦vcg to store data from channel 1') in the storage unit of the corresponding channel and store 7;
When an error is detected in the data stored by the corresponding channel after the dll unit has been cut, the date on which the error was detected is set to k [1 to keep all the corresponding addresses]. 'SA stack (Il-provided,
As a result, -4-
All addresses corresponding to all errors,
1! 4″l to be kept in the FSA stack
1. (5) Embodiment of the Invention The following is an error handling method in push-off control of a store operation as an embodiment of the present invention. I will explain it to you.

The error handling scheme according to the present invention can be applied in a system such as that shown in FIG. 1 as well as in the conventional case of Ail. FIG. 2 shows an example of the Ii'SA stack in the error handling scheme according to the present invention. In FIG. 2, the data to be written is BSK.

BS via input 1ull register BS・IN-REG
jiyt', the overflow data from the BS is written in
Output side register BS・0UT-1 (read through EG'i. 13 Address for loading into and reading from S is:

Address register ADI (specified by BE()). If an error occurs during write or read operations to BS, address register ADI (specified by BE()) The contents of are stored in the FSA stack (F8A-8TK) K read f i''I-te is held. If an error has occurred, the address for the newly generated error is read into the next area of the F8A stack.

Therefore, all a) pares corresponding to the error that occurred in 711J where FSA processing is performed are held in the I.'8A stack, and when processing F"SA, V is 11'.
Read all the same contents of SA stack -i:':t
Processing for L is performed.

Therefore, in Akishiko Kiichi's error handling method, for an error that occurs in the lack of store operation @1tL control: T'1 complete repair becomes possible, and system operation ff: continues execution. The monkey is the one who does it.

(6) Effect of the invention W According to the present invention, the channel and the channel peak II (,'til
, an error handling device that can handle the error that occurred when the leek 1 was disconnected is 111! and 13''c8).

[Brief explanation of the drawing]

Figure 1 is the push-off jtilJI1 created by StoreJTI.
f41: General configuration of /J・Iiwareruki1iV group 1cotff: Drag diagram. FIG. 2 is a diagram showing the error handling method in the control of the store operation according to the present invention. Patent applicant Fujitsu Co., Ltd. Patent application agent Patent attorney Akira Aoumi Patent attorney Nishidate 4 [1.Patent Attorney Yukio Uchida BUF Attorney Yama 1゛” Akiyuki

Claims (1)

Claims: A main storage unit, a buffer storage unit, a storage and channel control unit, a storage device with a plurality of channels, in which data from one of the channels is controlled. data in which the error was detected when an error related to the data stored by the channel was detected after the channel, the storage, and the channel control unit were separated when storing the channel to the main storage unit. F to hold the address corresponding to
The present invention is characterized in that an 8A stack is provided, and all addresses of data corresponding to all errors that occur until error handling is performed are held in the stack in F8. Error handling method in push-off control of store operation