JPH0814798B2 - Trace data reading system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] A trace data reading system in which trace data output from an information processing device is transferred to an external storage device and stored therein, and the presence of a lost event can be reliably detected. In order to make it possible to store the trace data output from the information processing device in the trace buffer storage device once and then transfer it to the external storage device, in the trace data reading system, It comprises a trace buffer pointer indicating how much trace data has been stored and a cycle counter indicating how many times the trace buffer storage device has been filled.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a trace data reading system in which trace data output from an information processing device is transferred and stored in an external storage device.

When a failure analysis or an operation analysis is performed when a hardware failure occurs in an information processing apparatus, trace data generated from the information processing apparatus is collected.
Then, failure analysis and operation analysis are performed based on the collected trace data. Here, the trace data refers to sequential recording of internal events and operation processes of hardware. Therefore, it is necessary to collect the trace data accurately in order to perform accurate analysis.

[Prior Art] FIG. 6 is a configuration block diagram of a conventional trace data reading system. The trace data output from the information processing device 1 is temporarily stored in the trace buffer storage device 2. Here, as the trace buffer storage device 2, a main storage device connected to a CPU (central processing unit),
A local memory or the like inside the system hardware is used. The stored trace data is read and transferred to the external storage device 4 such as a magnetic tape or a magnetic disk via the trace data transfer unit 3 and stored therein.

The trace data from the information processing device 1 is transferred to the external storage device 4 provided in the central processing unit (not shown) via the trace data transfer unit 3 each time it is generated. Here, if the storage speed of the external storage device 4 is faster than the frequency of generation of the trace data, the generated trace data is transferred to the external storage device 4 and stored as it is without providing the trace buffer storage device 2. Can be made.
However, when the storage speed of the external storage device 4 cannot follow the frequency of generation of trace data, the trace buffer storage device 2 for temporarily storing the trace data is provided as shown in FIG. It is stored in the storage device 2. When the trace buffer storage device 2 is full, a signal is generated from the trace buffer storage device 2 and given to the trace data transfer section 3 as an interrupt signal. When the trace data transfer unit 3 receives the interrupt signal, the trace data transfer unit 3 transfers the contents of the trace buffer storage device 2 to the external storage device 4 by the data transfer program. The trace buffer storage device 2 usually incorporates two or more trace buffers so that the trace data can be stored even during data transfer.

As described above, when the trace buffer storage device 2 becomes full, an interrupt signal is given to the trace data transfer unit 3, but if the trace data transfer unit 3 is operating with interrupts disabled, no interrupt will be accepted. . For this reason,
The trace data may be overwritten in the trace buffer in the trace buffer storage device 2 and the previous trace data may be lost. The trace data lost in this way is called a lost event.

The lost event will be described with reference to FIG. It is assumed that the trace buffer storage device 2 is composed of the trace buffers 1 and 2. These trace buffers shall have consecutive addresses from 1 to 2m. The trace buffer can store data from 1 to 2m. Therefore, 2m + 1 data is trace buffer 1
No. 1 is written. If the contents of trace buffer 1 are not transferred to the outside, 1 of trace buffer 1
Data will be sequentially lost from the second.

In order to perform accurate data analysis, it is necessary to reliably detect lost events. Therefore, the conventional lost event detection method uses time information (called time stamp) when the trace data is generated as shown in FIG.
Is written to the trace buffer as part of the trace data. Then, the lost event is detected by detecting the time difference between the time stamp of the first trace data and the 2m + 1th time stamp.

FIG. 9 is an explanatory diagram of the time difference detection method. The time difference can be detected by calculating the difference Δt (= T ₂ −T ₁ ) between the _first time stamp data T ₁ and the 2m + 1th time stamp data T ₂ . In this way, the lost event can be detected by detecting the time difference. That is, when the lost event has not occurred, the time difference has a constant value based on the trace data generation period. On the other hand, when the lost event occurs, the detected time difference Δt changes, so that the lost event can be detected. The lost event may be detected by a hand program in the trace data transfer unit 3 (see FIG. 6), added to the trace data and recorded externally, or detected at the time of analysis.

[Problems to be Solved by the Invention] The above-described method of detecting a lost event by a time stamp can be used only when the trace data is hardware trace data that occurs at a constant cycle.
When the trace data occurs aperiodically, the time stamp difference Δt is not constant even if the lost event does not occur. It is also conceivable that the difference Δt may take a constant value in spite of the lost event. Further, as shown in FIG. 10, when the time stamp makes one round at the address 2 m, there is a problem that the lost event cannot be detected even when the trace data is generated at a constant cycle.

The present invention has been made in view of the above circumstances, and an object thereof is to provide a trace data reading system capable of reliably detecting a lost event.

[Means for Solving Problems] FIG. 1 is a block diagram showing the principle of the present invention. The same parts as those in FIG. 6 are designated by the same reference numerals. In the figure,
Reference numeral 11 is a trace buffer pointer indicating to what extent trace data is stored in the trace buffer storage device 2,
Is a cycle counter indicating how many times the trace buffer storage device 2 is full. The trace buffer pointer 11 and the cycle counter 12 are the information processing device 1
The trace data output from the same is received in common and the output is given to the trace data transfer unit 3.

[Operation] The trace data output from the information processing device 1 is stored in the trace buffer storage device 2 and also given to the trace buffer pointer 11 and the cycle counter 12. The trace buffer pointer 11 updates its value each time the trace data is received. Cycle counter
12 receives the trace data and updates the data indicating how many times the trace buffer storage device 2 has been filled.
These data are sent to the trace data transfer unit 3. If the trace buffer pointer 11 points to the contents of the trace buffer that is about to be transferred, it will always be known that there is a lost event. In addition, if the value of the cycle counter 12 when the contents of the trace buffer were transferred last time is saved and compared with the current value of the cycle counter 12, the trace data transfer unit (in the central processing unit) 3 cannot interrupt. It is possible to detect whether or not all the trace data on the trace buffer has become a lost event in a certain period.

Further, the relationship between the trace pointer 11 and the cycle counter 12 will be described with reference to FIG. FIG. 11 is a diagram showing an example of a general configuration of the trace buffer storage device 2.
It consists of N trace buffers capable of storing trace data. The trace pointer 11 indicates the position of the last stored trace data, and every time the trace data is stored, the trace buffer 1 is 1, the trace buffer 1 is 2, ...
Trace buffer 1 m, trace buffer 2 1 and trace buffer 2 2 are sequentially updated to trace buffer N
When updated up to m, the trace buffer 1 is updated to 1. Trace pointer 11 is trace buffer N
, The cycle buffer 12 is updated because the trace buffer storage device 2 has been filled with trace data. As a result, the content of the cycle counter 12 represents the number of times the trace buffer storage device 2 has been filled with trace data.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a configuration block diagram showing an embodiment of the present invention. The same parts as those in FIG. 1 are designated by the same reference numerals. In the figure, reference numeral 20 is a central processing unit (CPU) that includes the trace data transfer unit 3 therein, and 30 is a main memory that includes the trace buffer storage device 2, the trace buffer pointer 11, and the cycle counter 12 therein. The trace buffer storage device 2 includes N trace buffers from the trace buffer 1 to the trace buffer N. Reference numeral 40 denotes a channel device, which includes therein a trace data collecting unit 41 that collects the trace data sent from the information processing device 1 and sends the trace data to the main storage device 30.
The operation of the apparatus configured as described above will be described below.

The trace data output from the information processing device 1 enters the trace data collection unit 41. The trace data collection unit
Reference numeral 41 sends the input trace data to the trace buffer storage device 2 and writes the trace data in the trace buffer, and updates the trace buffer pointer 11. Further, when it knows that the trace buffer is full, it notifies the central processing unit 20 of the trace buffer number and that the trace buffer is full. When the trace buffer N (last trace buffer) is full, the trace data collection unit
41 updates the cycle counter 12.

On the other hand, the trace data transfer unit 3 reads the trace buffer pointer value and the cycle count value from the trace buffer pointer 11 and the cycle counter 12, respectively,
The lost event is calculated and written in the external storage device 4, and then the trace data is read from the trace buffer of the notified trace data buffer number, and the channel device
Write to external storage device 4 via 40. As described above, according to the present invention, the lost event can be surely detected and written in the external storage device 4. Therefore, it is possible to accurately perform hardware failure analysis and operation analysis of the information processing device 1.

FIG. 3 is a flowchart showing the operation of the trace data collection unit 41. The trace data collection unit 41 first collects the trace data sent from the information processing device 1 (), and writes the collected trace data in the trace buffer (). Next, the trace buffer pointer 11 is updated (). Then, it is checked whether the trace buffer in which the trace data is stored is full (). When the trace buffer is full, the trace data transfer unit 3 is notified of the trace buffer number and the trace buffer being filled ().

Next, it is checked whether the trace number is N indicating the maximum number (). If it is the maximum number N, the cycle counter 12 is updated () and the process returns to the start. If the trace buffer is not full at step and the maximum number of trace buffers at step is N
If not, return to step.

FIG. 4 is a flowchart showing the operation of the trace data transfer unit 3. The trace data transfer unit 3 first reads the contents of the trace buffer pointer 11 and the cycle counter 12 (). Next, the difference between the internally stored cycle counter and the read cycle counter is calculated (). If the calculated difference is 1, the number of the trace buffer pointed to by the trace buffer pointer 11 is calculated by the following equation (however, the fractional part is truncated).

Y = {(content of trace buffer pointer 11-start address of trace buffer 1) / (length of trace buffer)} + 1 Next, the buffer number Y calculated from the trace buffer pointer 11 and the trace data collection unit 41 The buffer number X notified from is compared ().

When Y is the same as X, the following expression {(A−B) / C} −1 where A = trace pointer B = start address of trace buffer X C = value calculated by the number of bytes of trace buffer entry (lost event Value) to the external storage device (). When Y is smaller than X, zero is transferred to the external storage device as the value of the lost event (). When Y is larger than X, the following equation D / E, where D = trace buffer length E = value (lost event value) calculated by the number of bytes of the trace buffer entry is transferred to the external storage device ().

When the steps 1 to 3 are completed, all the trace data in the trace buffer X is transferred to the external storage device (),
Check if X has reached the final value N in the trace buffer (). When X becomes equal to N, the operation is terminated, and when not equal, the value of the cycle counter is stored inside the trace data transfer unit ().

FIG. 5 is a configuration block diagram showing another embodiment of the present invention. In the example shown, the trace buffer pointer
11 and the cycle counter 12 are provided in the local memory 50 instead of the main storage device 30. The read operation of the trace operation is the same as that of the embodiment shown in FIG. 2, and therefore its explanation is omitted.

[Effect of the Invention] As described above in detail, according to the present invention, the trace buffer pointer indicating how much trace data is stored in the trace buffer storage device and the number of times the trace buffer storage device has been filled are displayed. By providing the cycle counter shown, it is possible to provide a trace data reading system capable of surely detecting a lost event.

[Brief description of drawings]

FIG. 1 is a block diagram showing the principle of the present invention, FIG. 2 is a block diagram showing an embodiment of the present invention, FIG. 3 is a flow chart showing the operation of a trace data collection unit, and FIG. 4 is a trace data transfer unit. 5 is a flowchart showing the operation, FIG. 5 is a block diagram showing the configuration of another embodiment of the present invention, FIG. 6 is a block diagram showing the configuration of a conventional system, FIG. 7 is an explanatory diagram of a lost event, and FIG. 8 is a lost event detection. 9 is an explanatory diagram of a time difference detection method, FIG. 10 is an explanatory diagram of time difference detection, and FIG. 11 is a diagram showing a general configuration example of a trace buffer storage device. In FIG. 1, 1 is an information processing device, 2 is a trace buffer storage device, 3
Is a trace data transfer unit, 4 is an external storage device, 11 is a trace buffer pointer, and 12 is a cycle counter.

Claims (1)

[Claims] 1. A trace data reading system in which trace data output from an information processing device is temporarily stored in a trace buffer storage device, and then transferred to an external storage device for storage. A trace buffer pointer that indicates how much data has been stored, a cycle counter that indicates how many times the trace buffer storage device is full, and a trace data transfer unit that receives the output of these trace buffer pointer and cycle counter and calculates a lost event. A trace data reading system comprising: