JPH02304640A - Memory dump collecting system - Google Patents

Abstract

PURPOSE:To enhance the collectability of a memory image even by a small capacity by executing a memory dump collection of a process, and thereafter, executing the memory dump collection in order from the process in which a final execution time is near a system abnormality end generation time. CONSTITUTION:A memory dump collection file 2 is a file for collecting a memory at the system abnormality end generation time. A process execution time registration table 1 registers a final execution time of a process with regard to each process. At the time point when a CPU dispatch to each process is interrupted, a process execution time registering means 3 registers its time in a process execution time registration table 3. A memory dump collecting means 4 executes a memory dump collection with regard to the process which is being executed on a CPU at the system abnormality end generation time, and thereafter, refers to the table 1. Subsequently, the memory dump collection is executed in order from the process in which the final execution time is near the system abnormality end generation time, and at the time point when the file 2 becomes full, the collection of the memory dump is ended.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a memory dump collection method, and particularly to a memory dump collection method.
This paper relates to a memory dump collection method when a system terminates abnormally.

[Conventional technology]

As is well known, when an abnormal system termination occurs in a computer system, the operating system (
(hereinafter abbreviated as O8), a memory dump is collected as data for analyzing the cause of the system abnormal termination.

Conventionally, the following method has been used to collect memory dumps when a system terminates abnormally.

■ Collect all memory images of virtual memory (including temporary pages that do not exist in main memory).

■ Collect all memory images of main memory.By the way,
In the conventional memory dump collection method, no special consideration was given to the order in which memory images are flushed to the memory dump collection file. For example, in the method (■) above, the virtual address spaces are arranged in descending order of space number.
Within the space, memory dumps are collected in ascending order of intra-space addresses, and in the method (2) above, memory dumps are collected in descending order of real memory addresses.

[Problem to be solved by the invention]

In the conventional memory dump collection method described above, no special consideration was given to the order in which memory images are flushed to the memory dump collection file, so the size of the memory dump collection file on the disk device is small and the entire memory dump collection target is If the data cannot be stored, a memory image necessary for analyzing the cause of abnormal system termination cannot be obtained, and there is a drawback that it may be difficult to analyze the cause of abnormal system termination. In particular, as system scale tends to expand, it has become increasingly difficult to secure memory dump collection files of sufficient capacity, making the above drawbacks more prominent.

In view of the above-mentioned points, an object of the present invention is to make it possible to collect memory images necessary for analyzing the cause of abnormal system termination even if the capacity of the prepared memory dump collection file is small and it is not possible to collect the entire SOMORI dump. An object of the present invention is to provide a memory dump collection method that improves performance.

[Means for solving problems]

The memory dump collection method of the present invention is a method for collecting memory dumps when an abnormal system termination occurs in a computer system. a process execution time registration table for registering the final execution time of a process; a process execution time registration means for registering the time in the process execution time registration table when CPU dispatch to each process is interrupted; and system abnormal termination. After collecting memory dumps for the processes currently running on the CPU, the process execution time registration table is referred to, and memory dumps are collected in order from the processes whose last execution time is closest to the system abnormal termination time to create the memory dump collection file. and a memory dump collection means that finishes collecting the memory dump when the memory dump becomes full.

[Effect]

In the memory dump collection method of the present invention, the memory dump collection file is a file for collecting a memory dump when a system abnormal termination occurs, and the process execution time registration table registers the last execution time of each process, The process execution time registration means registers the time in the process execution time registration table when the CPU dispatch to each process is interrupted, and the memory dump collection means collects a memory dump of the process running on the CPU when the system terminates abnormally. After doing so, refer to the process execution time registration table and collect memory dumps in the order of processes whose last execution time is closest to the time when the system abnormal termination occurred, and stop collecting memory dumps when the memory dump collection file becomes full. .

〔Example〕

Next, the present invention will be explained in detail with reference to the drawings.

Note that the computer system to which the memory dump collection method of this embodiment is applied employs a multiple virtual memory method, and each virtual address space is classified into the following two types depending on its use.

+11 System space sharing module (O
3 cores) and a system-common O8 control table are stored.

(2) Process space: process-specific O8? l+I
I control tables and process specific code/data sections are stored.

FIG. 1 is a block diagram showing the configuration of a memory dump collection method according to an embodiment of the present invention. The memory dump collection method of this embodiment includes a process execution time registration table l,
Its main parts are composed of a memory dump collection file 2, a process execution time registration means 3, a memory dump collection means 4, and the memory 5 of the ta.

Memory dump collection file 2, which is a file for collecting memory dumps, is a file on the 6f'g disk device.

Referring to Figure 2, process execution time registration table 1
is a table whose index is the process number that uniquely identifies each process in the system, and within each entry, the corresponding process is finally
al Proc, essing [Jnlt] The time at which it was executed is stored.

Referring to FIG. 3, the process of the memory dump collection method of this embodiment includes the system space memory dump collection step 31.
, a process space memory dump collection step 32 for the running process, a process selection knob 33, a process space memory dump collection step 34 for the selected process, and a step 35 for determining completion of all process memory dump collection.

Next, the operation of the memory dump collection method of this embodiment configured as described above will be explained.

The process execution time registration means 3 includes a dispatcher (O3
It is incorporated as part of the core module (which performs cPU scheduling), and is incorporated as part of the CI (CI) of each process.
It operates when the U dispatch is interrupted (time slice expires, waiting for l10 (human output) completion, etc.), and writes the time at that point in the entry corresponding to the process in the process execution time registration table l.

On the other hand, the memory dump collection means 4 operates when the system terminates abnormally and collects the prepared memory dump collection file 2.
The image of the memory 5 at the time of abnormal system termination is copied to the memory dump collection file 2 within the range that can be stored in the memory 5 and so as to avoid losing the image of the memory 5 necessary for analyzing the cause of the system abnormal termination as much as possible. For this reason, the memory dump collection means 4 sequentially collects the virtual address spaces that are most likely to be necessary for analyzing the cause of system abnormal termination, that is, the system space, the process space of the process being executed at the time of system abnormal termination, and the last execution time. collects memory dumps from the memory 5 to the memory dump collection file 2 in the order of the process spaces of the processes closest to the system abnormal termination occurrence time.

Specifically, the memory dump collection means 4 first collects a memory dump of the system space from the memory 5 into the memory dump collection file 2 (step 31).

Next, the memory dump collection means 4 collects from the memory 5 a memory dump of the process space of the process that was being executed on the CPU (or each CPU in the case of the multiprocessor system) when the system abnormal termination occurred. Collect into file 2 (step 32).

Next, the memory dump collection means 4 refers to the process execution time registration table l, and selects one process from among the processes that were not being executed when the system abnormal termination occurred, in the order in which the last execution time is closest to the system abnormal termination occurrence time. Select (step 33) 0 Next, memory dump collection means 4
collects a memory dump of the process space of the selected process from memory 5 to memory dump collection file 2 (
Step 34). Subsequently, the memory dump collection means 4
It is determined whether the collection of memory dumps of all processes has been completed (Step 35), and if it has not been completed, control is returned to Step 33 to collect memory dumps of processes that were not running when the system abnormal termination occurred. The process from step 33 onward is repeated for processes for which the data has not been collected.

Then, when the collection of memory dumps for all processes is completed, or if the memory dump collection file 2 becomes full during memory dump collection (not shown in Fig. 3), the memory dump will be collected at that point. The collection means 4 ends the process.

In the above embodiment, a case has been described in which the memory dump collection target is limited to pages existing on the memory 5 of the main memory, but the present invention can be modified to collect memory dumps of entire virtual pages as necessary. It is also possible to implement.

〔Effect of the invention〕

As described above, according to the present invention, after collecting memory dumps for processes running on a cPU, the process execution time registration table is referred to, and memory dumps are sequentially stored starting from processes whose last execution time is closest to the time when the system abnormal termination occurred. By performing dump collection, even if the memory dump collection file prepared for memory dump collection when a system abnormal termination occurs is small and the entire memory dump cannot be collected, the memory required for analyzing the cause of the system abnormal termination can be saved. This has the effect of increasing the possibility of collecting images.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a memory dump collection method according to an embodiment of the present invention, FIG. 2 is a diagram showing the configuration of the process execution time registration table in FIG. 1, and FIG. 3 is the same as that shown in FIG. 1. 3 is a flowchart showing processing of a memory dump collection means in FIG. In the figure, l... Process execution time registration table, ゛ 2...
- Memory dump collection file, 3... Process execution time registration means, 4... Memory dump collection means, 5... Memory.

Claims (1)

[Claims] In a memory dump collection method when an abnormal system termination occurs in a computer system, a memory dump collection file that is a file for collecting a memory dump when an abnormal system termination occurs; A process execution time registration table that registers execution times, and C for each process.
a process execution time registration means that registers the time in the process execution time registration table when PU dispatch is interrupted; and a process execution time registration means that registers the time in the process execution time registration table when the PU dispatch is interrupted; A memory dump collection means that refers to a registration table, collects memory dumps in order of processes whose last execution time is closest to the time when the system abnormal termination occurred, and ends the collection of memory dumps when the memory dump collection file becomes full. A memory dump collection method comprising: