JP2000172578A - Fault processing system and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To restart the input/output instruction that is executed by an input/ output processor in the case of occurrence of a host bridge fault without ending abnormally the input/output instruction. SOLUTION: Each of three input/output processors 200 has an input/output instruction section 201 and a fault processing section 202. Every host bridge 100 is placed in 1:1 to every processor 200 and controls the interface between a CPU 400 and the processor 200. A fault notification means 103 notifies the processor 200 of the fault of the bridge 100. Every the fault processing section 202 collects and analyzes the information on the fault of the bridge 100 and recovers this bridge. Every the instruction section 201 stops temporarily the input/output instruction under execution and then restarts it after executing the fault processing by the fault processing section 202 and the bridge 100 is recovered.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer fault handling system, and more particularly, to a fault handling in an input / output processing device against a host bridge fault.

[0002]

2. Description of the Related Art FIG.
Shown in 5 and 6 show the operation of the conventional fault handling system.

In FIG. 4, three host bridges 10
0 indicates each host bridge interface 10
1 through a central processing unit (hereinafter referred to as a host) 400
And a diagnostic control device 300 via a diagnostic interface 301 and an input / output command processing device 203 via an input / output processing device interface 104, respectively.
Further, each host bridge 100 and the input / output processing device 203
Are connected to the diagnostic control device 300 by the fault notifying means 103.

A normal input / output command is issued by an operating system or application software running on the host 400, and the input / output command is transmitted to the host bridge 100 via the host bridge interface 101. The host bridge 100 having received the input / output command transmits the input / output command to the input / output processing device 203 connected thereunder via the input / output processing device interface 104. The input / output processing device 203 that has received the input / output command activates the input / output command processing unit 201, and the input / output command processing unit 201 executes the input / output command.

Data transfer is performed to the host 400 via the input / output processing device 203, the input / output processing device interface 104, the host bridge, and the host bridge interface 101 in the case of a read command. Then, the data of the host 400 is transferred to the host bridge interface 101, the host bridge 100, and the input / output processing device interface 1.
The data is transferred to the input / output processing device 203 via the input / output device 04.

At this time, the transfer data is divided into a certain amount and transferred a plurality of times. Operating system or application software and input / output command processing unit 20
In this case, the data transfer executed a plurality of times is executed in response format, and when all the data transfer is completed, the input / output command processing unit 201 issues a report of the end to the operating system and the data transfer ends.

The conventional fault handling system for a computer system comprises a host bridge 100 and an input / output device 20.
The fault of No. 3 is reported to the diagnostic processing device 300 via the fault notifying means 103 and processed by the fault processing unit 302.
In the fault handling system having this configuration, when a fault occurs in the host bridge 100 during the data transfer sequence described above, the host bridge 100 reports its own fault to the diagnostic control device 300 via the fault notifying means 103, and Failure processing is performed by the unit 302.

When a failure occurs in the host bridge 100, the input / output instruction executed by the input / output instruction processing unit 201 of the input / output processing unit 203 connected thereunder contradicts (the data transfer is executed. No response is returned)
Is detected, the input / output instruction cannot be continuously executed, and the input / output instruction processing unit 201 detects an input / output instruction exception. The I / O processing device 203 that has detected the I / O instruction exception reports the I / O instruction exception to the diagnosis control device 300 via the failure notification unit 103, and is processed by the failure processing unit 302 as the I / O processing device exception.

Further, when a failure occurs in a plurality of host bridges 100 at almost the same time, the failure notification of all the host bridges 100 and all the input / output processors 2
The input / output instruction exception notification 03 is reported to the only diagnostic processing device 300 in the system via the failure notification means 103 and is processed by the failure (exception) processing unit 302.

[0010]

However, the conventional fault handling system described above has the following problems.

The first problem, as shown in FIG. 5, is that the execution of an input / output instruction in the input / output instruction processing unit 201 of a sound input / output device 200 connected to the host cannot be performed due to a host bridge failure. This causes an abnormal end of the input / output instruction (input / output instruction exception). This is because the input / output command processing unit 201 of the input / output processing device 200
Recognizes the failure state of the host bridge 100 as an inconsistency during execution of an input / output instruction, and sends a failure notification to the failure notification means 103.
To the diagnostic control device 300 via the I / O instruction processing unit 201.
This is because the operation cannot be performed until the re-initialization processing from the failure processing unit 302.

A second problem is as shown in FIG.
When a plurality of host bridge 100 failures occur almost simultaneously, the load on the entire failure processing system increases, and it becomes difficult to quickly perform the failure processing. This is because the fault processing of all the host bridges 100 and all the input / output processing devices 200 is performed by the fault (exception) processing unit 302 of only one diagnostic control device 300 in the system.

Accordingly, an object of the present invention is to provide a failure notification means for reporting a failure of a host bridge to an I / O processing device under the host bridge, and a failure processing unit for a failure of the host bridge in the I / O processing device. As a result, the I / O instruction executed by the I / O processing unit at the time of occurrence of the host bridge failure is restarted without abnormal termination, and the load in the failure processing system is distributed to execute the failure processing in parallel. It is an object of the present invention to provide a fault handling system capable of performing the task quickly.

[0014]

The fault processing system according to the present invention comprises at least one input / output processing unit having an input / output instruction processing unit for processing input / output instructions from an operating system or application software;
In a fault processing system in a computer system having a one-to-one correspondence host bridge with the input / output device controlling an interface between a host and the input / output processing device, a fault notification for notifying the input / output processing device of a host bridge fault Means and a failure processing unit for collecting and analyzing failure information of the failed host bridge and performing recovery of the host bridge are provided for each of the input / output processing devices, and a failure occurs in the host bridge and the input / output instruction cannot be executed. The host bridge fault is reported to the input / output command processing unit and the fault processing unit of the input / output processing device via the fault notification unit, and the input / output command processing unit After the execution of the fault processing by the fault processing unit and the recovery of the host bridge, the suspended I / O instruction is re-executed. Characterized in that to.

Further, according to the fault processing method of the present invention, at least one input / output processing device including an input / output command processing unit for processing input / output commands from an operating system or application software, a host, and the input / output processing device A failure handling method in a computer system having the input / output device controlling an interface between the input / output device and a host bridge having a one-to-one correspondence, a procedure for notifying the input / output processing device of a host bridge failure;
A procedure in which the input / output processing device that has received the notification executes a procedure for collecting and analyzing fault information of the failed host bridge, and restoring the host bridge, and a failure occurs in the host bridge, and the input / output instruction cannot be executed. In this case, a procedure for reporting the host bridge fault to the input / output command processing unit and the fault processing unit of the input / output processing device via the fault notification means, and the input / output command processing unit receiving the report, The method further comprises a step of suspending the instruction, and restarting the suspended input / output instruction after executing the failure processing by the failure processing unit and restoring the host bridge.

More specifically, in FIG. 1, a failure that has occurred in the host
Each of the input / output processing devices 200 has a failure notifying unit 103 for reporting to the I / O processing device 200 and a failure processing unit 202 for performing a failure process of the host bridge 100 connected to the input / output processing device 200.

That is, a failure notifying means 103 for notifying a host bridge failure is provided between each host bridge 100 and the input / output processing device 200.
0 is provided with a failure processing unit 202 for performing failure processing of the host bridge 100.

Therefore, when a host bridge fault occurs, the fault processing unit 202 of the input / output processing unit 200 under the control of the host bridge fault
It is possible to know that the currently executed input / output instruction cannot be continuously executed due to the occurrence of a failure in the host bridge 100, and the input / output instruction executed by the input / output instruction processing unit 201 of the input / output processing device 200 The input / output instruction that has been executed can be temporarily stopped, and the input / output instruction can be re-executed after the reinitialization (restoration) processing of the host bridge 100.
The system operates without abnormally terminating the I / O command being executed, and the failure of a plurality of host bridges 100 causes the I / O processor 20 connected to each
0, the failure processing units 202 can perform processing in parallel, and the failure processing of each host bridge 100 can be performed quickly.

[0019]

Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows an embodiment of a fault handling system according to the present invention. In FIG. 1, three host bridges 10
0 indicates each host bridge interface 10
1 and the input / output instruction processing device 20 via the input / output processing device interface 102.
Connected to 0. Also, each host bridge 10
0 indicates that the input / output processing device 200
Connected to.

The host bridge 100 transmits an input / output command issued from an operating system or application software to the input / output processing device 200 via the interface control unit 102. In the input / output processing device 200 that has received the input / output instruction, the input / output instruction processing unit 201 executes the input / output instruction and executes input / output processing such as data transfer.

In the case of a read command, data transfer is performed to the host 400 via the input / output processing device 200, the input / output processing device interface 102, the host bridge 100 and the host bridge interface 101, and the write command is transmitted. If so, host 400
Is transmitted to the input / output processing device 200 via the host bridge interface 101, the host bridge 100, and the input / output processing device interface 102.

At this time, the transfer data is divided into a certain amount and transferred a plurality of times. Operating system or application software and input / output instruction processing unit 2
01, the data transfer executed a plurality of times is executed in a response format, and when all the data transfer is completed, the input / output command processing unit 201 issues an end report to the operating system and the data transfer ends.

When a failure occurs in the host bridge 100 during data transfer, the present system reports this failure to the I / O processing unit 20 connected under the host bridge 100.
0, a failure notification means 103 for reporting to
The input / output processing device 200 includes a failure processing unit 202 that can collect and analyze failure information of the host bridge 100 and re-initialize (restore) the host bridge 100;
Input / output instruction processing unit 2 when host bridge 100 failure occurs
01 is provided with an input / output instruction processing unit 201 that can execute a pause and a re-execution process of the input / output instruction that has been executed.

Next, the operation of the fault handling system will be described with reference to FIGS.

FIG. 2 shows a sequence when a failure occurs in the host bridge 100 during execution of an input / output command (data transfer command).

When a failure occurs in the host bridge 100 during execution of an input / output command issued from the operating system or application software, the host bridge 100 reports the occurrence of its own failure and notifies the failure notification means 10
3 to the input / output processing device 200.

The I / O processing unit 200, which has been notified of the occurrence of the failure of the host bridge 100, determines that the I / O instruction currently being executed by the I / O instruction processing unit 201 cannot be continuously executed. To notify the host bridge failure occurrence that the currently executed input / output instruction cannot be continuously executed.

Upon receiving the host bridge fault notification, the input / output command processing unit 201 suspends the currently executing input / output command and waits for an input / output command re-execution instruction from the fault processing unit 202. The input / output instruction processing unit 201 that has temporarily stopped the input / output instruction activates the failure processing of the failure processing unit 202.

The fault processing unit 202 started from the input / output command processing unit 201 is a
0 failure information (log data) is collected via the input / output processing device interface 102. The collected failure information is analyzed by the failure processing unit 202. Failure processing unit 201
When it is determined in the failure information analysis that the host bridge 100 can be continuously operated by reinitializing the host bridge 100, the failure processing unit 202
Is reinitialized. In this re-initialization processing, the failed state of the host bridge 100 that has failed is cleared, and the interface between the host 400 and the input / output processing device 200 can be operated again.

After the reinitialization of the host bridge 100 is completed, the fault processing unit 202 notifies the input / output command processing unit 201 that the host bridge 100 has been restored to the operable state due to the recovery. To re-execute the input / output instruction at the time. Upon receiving the input / output instruction re-execution instruction, the input / output instruction processing unit 201 becomes unexecutable when the host bridge failure occurs, resumes the temporarily stopped input / output instruction, and returns to the normal data transfer sequence.

Next, FIG. 3 shows a plurality of host bridges 10.
A case where a failure occurs almost simultaneously at 0 will be described.

All the failed host bridges 10
0 notifies a fault to the subordinate input / output processing device 200 via the respective interface control units 102 and the fault notification means 103. Each input / output processing device 200
Input / output command processing unit 2 of each host bridge 100
01, the input / output instruction being executed is temporarily stopped.
1 to perform failure processing for only the host bridge 100 to which it is connected. After the reinitialization processing of the self-connected host bridge 100 by the failure processing unit 202 is completed, the input / output instruction processing unit 201 enables the input / output instruction that has become unexecutable at the time of the host bridge failure to be executable again.

[0034]

According to the present invention, the following effects can be obtained by adopting the configuration as described above.

When the host bridge failure occurs, the input / output instruction being executed in the input / output instruction processing unit of the input / output processing device is temporarily stopped, and the input / output instruction is re-executed after reinitialization in the host bridge failure processing. Accordingly, the system can be operated without causing an exception (abnormal termination) to the input / output instruction being processed, and the fault processing units of the subordinate input / output processing devices can process a plurality of host bridge faults in parallel. Therefore, it is possible to quickly execute each host bridge failure process.

[Brief description of the drawings]

FIG. 1 is a block diagram of an embodiment of a computer system fault handling system according to the present invention.

FIG. 2 is an operation flowchart at the time of a host bridge failure in the failure processing system of the computer system according to the present invention;

FIG. 3 is an operation flowchart at the time of simultaneous failure of a plurality of host bridges in the failure processing system of the computer system according to the present invention;

FIG. 4 is a block diagram of a conventional computer system fault handling system.

FIG. 5 is an operation flowchart at the time of a host bridge failure in a conventional computer system failure handling system.

FIG. 6 is an operation flowchart for a simultaneous failure of a plurality of host bridges in a failure processing system of a conventional computer system.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 host bridge 101 host bridge interface 102 input / output device interface 103 fault notification means 104 input / output processing device interface 200 input / output processing device (IOP) 201 input / output command processing unit 202 fault processing unit 203 input / output processing device (IOP) 300 diagnosis Control device 301 Diagnostic interface 302 Fault processing unit 400 Host

Claims (2)

[Claims] 1. An at least one input / output processing unit having an input / output instruction processing unit for executing input / output instruction processing from an operating system or application software, and an interface between a central processing unit and the input / output processing unit. A fault notification system for notifying a host bridge fault to the input / output processing device, and fault information of the faulty host bridge. A fault processing unit for collecting, analyzing, and restoring the host bridge is provided for each of the input / output processing devices. If a fault occurs in the host bridge and the input / output instruction cannot be executed, the host bridge fault is regarded as the fault. The input / output command processing unit of the input / output processing device and the The input / output instruction processing unit reports the processing to the processing unit. The input / output instruction processing unit temporarily suspends the currently executing input / output instruction, and resumes the temporarily stopped input / output instruction after performing the failure processing and restoring the host bridge in the failure processing unit. A fault handling system characterized by causing a fault. 2. An at least one input / output processing device having an input / output instruction processing unit for executing input / output instruction processing from an operating system or application software, and before controlling an interface between a host and the input / output processing device. In a failure processing method in a computer system having an entry / output device and a one-to-one host bridge, a procedure for notifying a host bridge failure to the input / output processing device; The procedure for collecting, analyzing, and restoring host bridge fault information that has been lost, and, if a host bridge fault occurs and I / O commands become unexecutable, report the host bridge fault via the fault notification means. Reporting to the entry / exit instruction processing unit and the failure processing unit of the entry / exit processing device, A procedure in which the input / output instruction processing unit receiving the report temporarily suspends the currently executing input / output instruction, and resumes the suspended input / output instruction after execution of the failure processing by the failure processing unit and restoration of the host bridge. And a failure handling method.