JPH0442695B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a method for creating event trace data in a computer system, and in particular, a method for creating event trace data that is effective for failure analysis of input/output devices in a computer system. Regarding the method.

[Conventional technology]

In computer systems, the internal state and time of events such as the occurrence of an interrupt between devices or the start of a program are recorded as event trace data, and in the event of an abnormality, this data is read out and used for operational analysis. It will be done. The conventional method for creating such event trace data is
59-111552, 59-223864, etc., the occurrence of control data such as program start, termination, and interrupts that can be easily collected in a host computer (hereinafter simply referred to as the host), measured by a timer in the host. Events of input/output control devices connected to the host were recorded independently along with the time of occurrence measured by the device's own timer.

[Problem that the invention seeks to solve]

In the above conventional technology, if the input/output control device connected to the host is a highly functional one, for example, one that uses a DMA control method as an interface with the host, or one that is equipped with a microcomputer, the operation of the host program It is difficult to correlate the timing of access to the input/output control device with the internal operation of the input/output control device and the timing of information transmission because it is measured by an independent timer. Trouble analysis in the event of an abnormality and integration tests with terminals connected to the input/output control device required a great deal of effort, and it was also necessary to add dedicated monitoring tools.

On the other hand, in order to deal with these problems, there are methods in which the event trace function in the input/output control unit directly reads the date and time information given by the timer in the host and records trace data, and Methods of inquiring about the status of the I/O controller may also be considered, but these methods would cause a decline in the performance of the I/O operations that the host or the I/O controller should normally perform, and would also cause problems with the operation of the I/O controller. The disadvantage is that it becomes complicated.

The present invention can add time information synchronized with the host's timer to event trace data of an input/output control device connected to a host, and this does not reduce the performance of the host or input/output control device. The purpose is to provide a method for creating event trace data without the need for

[Means for solving problems]

The purpose of the above is to provide a timer in each of the host and input/output control device as in the past, record the time indicated by the timer of each device as the event occurrence time, and trace the event by starting the event trace based on commands from the host. The input/output control device is provided with a start means and a reset means for resetting the built-in timer, and when the host commands the input/output control device to start tracing, it stores the start time indicated by the timer in its own device at that time, and then , when the event trace data recorded by the input/output control device is read, the stored start time is added to the occurrence time of each read event, and is achieved by making the event occurrence time, Furthermore, when the host receives a message containing time data measured by the terminal's timer from a terminal connected via an input/output control device, it converts the terminal's time data at the time of reception to the time determined by the host timer. This is achieved by determining the difference between the values and correcting the terminal's timer count accordingly.

[Effect]

By the above trace start and set means, the timer in the input/output control device is reset to 0 at the start of event tracing, and the time of the host timer at that point is later added to the trace data, so the input/output control device timer is reset to 0 when the event trace starts. Since the event occurrence time within the input/output control device can be measured almost accurately using the timing of the host timer, no extra burden is placed on the processor. Furthermore, when receiving a time data message from a terminal, the timing of the terminal side timer is synchronized based on the time error at the time of reception.
Binding tests can be performed easily.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG. In this diagram, host 1 is replaced by input/output control device 2.
is connected via an interface 2e, and information and control commands in a random access memory (RAM) 2d are transferred. Also interface 2
Terminal 3 is connected via f, and data transfer is also performed with this terminal. The host 1 includes a main memory 1a in which programs and data are stored, a processor 1b that executes the programs, and a date and time counting mechanism 1 that can be read from the processor.
It consists of c. The input/output control device 2 decodes the input/output command from the host 1, outputs an interrupt signal INT0 to the microprocessor 2a, and
An interface 2e that transfers data to and from the computer, a random access memory 2d, a memory (ROM) 2c that stores a program, a microprocessor 2a that executes this program, a timer 2d that generates an interrupt signal INT2 with a 1ms cycle, and a terminal. It consists of an interface 2f with 3. The interface 2f sends an interrupt signal to the microprocessor 2a every time data transmission/reception with the terminal is completed.
Output INT1.

FIG. 2 shows the configuration of an event trace data storage area 2d1 in the random access memory 2d. Case No. in the same area indicates the number of event trace data (number of cases) stored in order from address A+2, and a pair of data (i) and counter (i) is displayed as one case. The address pointer at address A+1 indicates the address where trace data is to be stored next, as indicated by the dotted line in the figure. In area 2d2, the count value of timer 2b is B.
+0 address, event data is stored at address B+1, and a flag indicating that a trace start command has been issued from the host is stored at address B+2. Counter (i) and data (i) are stored in one case at the address indicated by the address pointer in area 2d1 when an event occurs.

3 and 4 are programs for event tracing stored in the ROM 2c.
In FIG. 3, the timer interrupt processing program 2c
1 is activated by the interrupt INT2 from the timer 2b with a 1 ms cycle, and updates the counter value at address B+0 in FIG. 2 by +1.

The host command processing program 2c2 is started by the interrupt INT0 from the host, and if this interrupt is a trace command (step 200), it resets address B+0 in the area 2d2 in FIG. (corresponding to resetting the timer of the output control device), instructs the event trace function to start tracing (step 202), and further resets the case number of area 2d1 to 0.
On the other hand, when the interrupt INT0 is an input/output control command, data transfer between the main memory 1a and RAM 2d of the host is controlled (step 204).

The event occurrence processing program 2c3 is the terminal 3
This is activated by the interrupt INT1 from the interface 2f when data is exchanged with the interface 2f, and an event trace function (subroutine) is activated.

The event trace function 2c4 in FIG. 4 is activated by the event generation process 2c3, and if the trace start flag is turned on in the step of the host command process 2c2 by a trace start command from the host (step 400), the event trace function 2c4 is activated by the event generation process 2c3. Figure 2 area 2d1
Check Case No. (Step 401). If this is not the maximum, add 1 to Case No. (step 402),
B to the address pointed to by the address pointer in area 2d1
Store the timer count value at address +0 (step
403), stores the data at address B+1 to the next address (step 404), increments the address pointer by 2 (step 405), and ends the process if the address pointer has not reached its maximum value A+N (step 406). If the value has reached the maximum value, the value of the address pointer is set to address A+2 in preparation for the next event trace (step 407).

FIG. 5 shows the data storage area in the main memory 1a of the host 1. When the event trace start command is output, the date and time data 1a1 (address L) indicated by the internal host timer 1c, and the data stored in the received message from the terminal 3 are shown in FIG. Date and time data 1a included in
2 (address M), and date/time data 1a3 of host timer 1c when receiving data from this terminal.
(address K) and event trace data in RAM 2d of input/output control device 2 (area 2d in Figure 2).
It consists of area 1d4 where 1) is read and stored.

FIG. 6 shows a program stored in the main memory 1a of the host 1. The event trace start processing 1a5 includes the date and time of the host timer 1c at the time of start.
Read the time and store it at address L in Figure 5 (step 500), and output a start command (step 500).
501).

Transfer the event trace data 2d1 to the main memory area 1a4 (Fig. 5) (step 601),
Check whether to synchronize with the host's timer or the terminal's time (step 602. This selection depends on the operator's settings), and when host synchronization is performed, the counter (i) in area 104 is set at address L in Figure 5. The time is added (step 603), and when the terminal is synchronized, the difference in time between addresses M and K in Figure 5 is calculated and stored in address L (step 604), and then this value is stored in the counter (i) of area 104. Add (step
603).

The terminal data reception processing program 1a7 extracts the date and time information in the received data from the terminal 3 and stores it at address M in FIG. 5 (step 700), and sets the date and time of the host timer 1c at that time in FIG. K
Store it at the address (step 701).

The above is the detailed configuration of the embodiment of the present invention,
Their overall operation is as follows. host 1
In the case of event tracing in the input/output control device 2, the event trace command processing program 1a5 shown in FIG. Interrupt INT0 is turned on. Then, the host command processing 2c2 shown in FIG. 3 is activated, and an event tracer start command (step 202) is issued here. Note that the timer interrupt processing program 2c1 is always in operation upon system startup. On the other hand, the event generation processing 2c3 (Fig. 3) in the input/output control device 2 is activated by the interrupt INT1 during data transmission from the terminal 3, thereby starting the program shown in Fig. 4, and sequentially transmitting the event trace data to the area. 2d1. Thereafter, when receiving a command from the host operator or when an abnormality occurs, the event trace data real-time correction program 1a6 is executed, and time correction is performed on the trace data in synchronization with the time of the host timer.

〔Effect of the invention〕

As is clear from the above embodiments, according to the present invention, it is possible to detect the time course of events of input/output control devices and terminals connected to the host in synchronization with the timer in the host, so that abnormalities in the computer system can be detected. It has the effect of speeding up operation analysis and making it easy to perform connection tests with terminals. Moreover, for this purpose, it is only necessary to provide a simple timer in the input/output control device, and for synchronization, input/output There is no significant burden placed on the control device or host processor.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention;
Figure 2 is an explanatory diagram of event trace data and timer area in the input/output control device, Figure 3 is a flowchart of the input/output control unit program, Figure 4 is a flowchart of the event trace function, and Figure 5 is the memory in the host. FIG. 6 is a flowchart of the host program. 1...Host, 2...I/O control device, 3...
Terminal, 1a...Host memory, 1c...Date and time counting mechanism, 2b...Timer, 2d1...Event trace data, 2d2...Timer counting counter, 2c2...Host command processing, 2c4...Event trace function, 1a5 ...Event trace command processing, 1a6...Real-time correction processing of event trace data.

Claims (1)

[Scope of Claims] 1. When the host computer outputs an event trace start command, the time indicated by the built-in clock means is stored as the initial time, and the input/output control device that receives the start command starts the timer in its own device. After resetting, the timer starts counting and moves to the event trace state, and if an event occurs in that state, the event data indicating the event and the count value of the above timer at the time of occurrence are stored as event data in the built-in memory. The host computer reads the contents of the event data area in the memory of the input/output control device when an instruction from the operator or abnormality is detected, and sets the timer count value in the read data to the time indicated by the value. A method for creating event trace data, characterized in that the event trace data is replaced with a time added to the initial time stored at the time of the start command. 2. When data including the current transmission time measured by the timer on the terminal side is transmitted to the host computer from a terminal connected to the input/output control device, the host computer Claim 1, characterized in that the difference between the time indicated by the built-in timer and the transmission time is calculated and used as an initial time in place of the initial time stored at the time of the start command. How to create event trace data.