Method for positioning faulted memory in linux system

Abstract

The invention provides a method for positioning a faulted memory in a linux system. The method comprises that error types and fault positions of the faulted memory are judged quickly and problems such as server halting and blue screens caused by memory faults are solved quickly by usage of mcelog records produced by the system according to the actual physical slot comparison relationship without regard to judgments and records of memory faults provided by a mainboard baseboard management controller (bmc). Compared with existing memory fault judgment methods, the method has the advantages that the method is independent of a memory fault tracking and positioning function of the mainboard bmc; on-site repetition of pressure tests to wait for faults and exchange of memory tests to position the faulted memory are not required, and the fault appearance frequency, fault reasons and fault positions can be determined quickly according to generated mcelog logs; and the implementation method is simple, and the efficiency for judgment and solving of problems can be improved.

Description

technical field [0001] The invention relates to the field of computer applications, in particular to a method for locating fault memory locations under a linux system. Background technique [0002] For the judgment of the faulty memory location, there are two general methods before: [0003] 1) Using the memory fault location and recording function of the motherboard integrated bmc, when a correctable ecc error or an uncorrectable ecc error occurs in the memory, the motherboard bmc can record the memory error information and record the memory slot where the fault occurred, so as to quickly locate the faulty memory location , but the use of this method has certain limitations. First, it must be ensured that the server has a bmc management chip. However, bmc management is a technology that has only been used in recent years. Early general models do not have a bmc management chip; moreover, even if The machine comes with a bmc chip, which does not necessarily have the function...

AI Technical Summary

Problems solved by technology

[0003] 1) Using the memory fault location and recording function of the motherboard integrated bmc, when a correctable ecc error or an uncorrectable ecc error occurs in the memory, the motherboard bmc can record the memory error information and record the memory slot where the fault occurred, so as to quickly locate the faulty memory location , but the use of this method has certain limitations. First, it must be ensured that the server has a bmc management chip. However, bmc management is a technology that has only been used in recent years. Early general models do not have a bmc management chip; moreover, even if The machine comes with a bmc chip, which does not necessarily have the function of memory fault location and needs to be developed independently, so it is still not possible to realize the location and detection of memory faults; the bmc memory fault location function is limited by the existence of the management chip and its own functions, and cannot become a general solution;

[0004] 2) Memory stress test: After the initial judgment that it is a memory fault, it is necessary to locate which memory has a fault from more than a dozen memory, use the memory stress test tool to test the system platform and memory in batches, gradually narrow the scope, and finally locate There are certain limitations in the use of this method for any memory failure: First, for the stress test of large-capacity memory, the failure recurrence time is not easy to grasp, and it is very likely that the stress test may not necessarily reproduce the failure in one day, especially for customers who have a failure about once a week It is more difficult to reproduce the problem on the machine; second, the overall test time is too long when using the method of batch testing; third, it cannot be ruled out that it is not possible to accurately locate the problem due to poor memory contact or a problem with the memory controller of the CPU itself;

Images