A memory fault handling method, computing device and management platform
By statistically analyzing and jointly determining the risks of storage units and their associated areas in memory modules, the problem of inaccurate memory fault isolation is solved, the accuracy of memory fault handling is improved, and the risk of system downtime is reduced.
Patent Information
- Application Number
- CN202410119037.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-29
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2044-01-29
Smart Images

Figure CN117971539B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a memory fault handling method, computing device and management platform. Background Technology
[0002] Memory is a crucial component of computing devices, and memory failure is one of the most common hardware failures in computing systems, significantly impacting the reliability, availability, and performance of the entire system. Specifically, when memory errors occur, the memory controller can employ error checking and correction (ECC) algorithms to correct them. However, frequent error correction can negatively affect system performance, and as memory failures become more severe, the error correction algorithms become overwhelmed, significantly increasing the probability of errors becoming uncorrectable. This can lead to uncorrected errors (UCEs), ultimately causing the computing device to crash.
[0003] To prevent system downtime from impacting business operations on computing devices, related technologies typically isolate memory locations where frequent errors occur, thus avoiding further read / write operations on those memory locations and reducing the probability of uninterrupted memory errors (UCEs). However, this isolation method has low accuracy and may not be able to completely prevent system downtime caused by memory failures. Summary of the Invention
[0004] This application provides a memory fault handling method, computing device, management platform, computer storage medium, and computer program product, which can improve the accuracy of isolation of faulty memory regions.
[0005] In a first aspect, embodiments of this application provide a memory fault handling method, the method comprising: acquiring fault information, the fault information including address information and fault time of a faulty storage unit, the faulty storage unit being a storage unit in memory where a correctable error (CE) has occurred, wherein the memory includes at least one memory column, each memory column including multiple memory particles, and each memory particle including several storage units; determining the risk status of the faulty storage unit and the risk status within associated storage particles based on the fault information, wherein associated storage particles are memory particles located in the same memory column as the memory particle to which the faulty storage unit belongs, wherein the risk status of the faulty storage unit includes the risk level of the faulty storage unit and the risk level of the memory region surrounding the faulty storage unit, the risk level being used to represent the degree of risk of an uncorrectable error (UCE) occurring, and the risk status within associated storage particles including the risk level of their own internal storage units; and determining a risk area of the memory based on the risk status of the faulty storage unit and the risk status within associated storage particles, wherein the risk area of the memory is an area that needs to be isolated.
[0006] In this embodiment, memory chips within the same memory rank on a memory module share a memory channel. During read operations, data is read from the same location (i.e., the same level of storage array (bank), the same row address, and the same column address) of multiple memory chips (devices) within the same memory rank. Therefore, the risk of a faulty memory chip interacts with the risk of its internal storage units (primarily the cells within the same location as the faulty chip and other cells surrounding it). Furthermore, the risk level of a single storage unit is also influenced by the risk levels of its surrounding storage units. Consequently, if the risk levels of the faulty memory chip, its surrounding storage units, and the internal storage units of its associated memory chips are all high, the risk of a UCE (Unified Risk Emission) will be significantly increased. Therefore, this embodiment combines the risk levels of the faulty memory chip and its surrounding area, as well as the risk levels of the internal storage units of its associated memory chips, to perform a joint risk assessment across chips, identify risky areas for isolation, and improve isolation accuracy.
[0007] In some possible implementations, the risk status of the faulty storage unit is determined based on the fault information, including: querying the historical risk records of the faulty storage unit and the surrounding memory region from the fault risk database based on the fault information, wherein the historical risk records include historical risk levels; and updating the risk level of the faulty storage unit and the risk level of the surrounding memory region based on the historical risk records of the faulty storage unit and the surrounding memory region to obtain the risk status of the faulty storage unit.
[0008] In this example, since a CE (Complete Error) in one cell implicitly increases the failure risk of surrounding cells, this example maintains a failure risk database to record the historical risk records of all cells in all memory chips. When a faulty memory cell occurs, the historical risk level of the faulty memory cell and its surrounding area can be queried and updated from this failure risk database, which helps to improve the accuracy of subsequent risk area determination.
[0009] In some possible implementations, each memory chip includes a multi-layer storage array, and each layer of the storage array includes several storage cells arranged in a matrix. The risk level of the faulty storage cell and the risk level of the surrounding memory region are updated to obtain the risk status of the faulty storage cell, including: within a preset period, if the faulty storage cell has never experienced a CE (Collision Error) in its history, and the number of CEs and / or the number of storage cells experiencing CEs in the row, column, and storage array where the faulty storage cell is located do not reach the preset risk conditions, then the risk level of the faulty storage cell is determined to be Level 1; within a preset period, if the faulty storage cell has experienced a CE in its history, and the number of CEs and / or the number of storage cells experiencing CEs in the row, column, and storage array where the faulty storage cell is located do not reach the preset risk conditions, then the risk level of the faulty storage cell is determined to be Level 2, where the risk level of Level 2 is higher than that of Level 1.
[0010] In this way, based on the updates to the historical risk records of the faulty storage unit itself and the historical risk records of the surrounding cells, the impact of the current faulty storage unit on itself and the surrounding cells when this CE occurs is digitized, so as to provide reliable data reference for subsequent accurate determination of risk areas.
[0011] In some possible implementations, determining the risk status of a faulty storage unit based on fault information also includes: if a faulty storage unit's row has historically experienced a CE (Critical Error Correction) within a preset period, then all storage units in that row that have experienced a CE are classified as Level 2 in terms of risk level, and all storage units in that row that have not experienced a CE are classified as Level 1 in terms of risk level, and that row is marked as a risky row.
[0012] In this example, when the current faulty storage cell experiences this CE, it affects the risk status of other cells in its row. Therefore, marking the risk level makes it easier to accurately determine the risk area later.
[0013] In some possible implementations, determining the risk status of a faulty storage unit based on fault information also includes: if a faulty storage unit has experienced a CE (Critical Error) in the column containing the faulty storage unit within a preset period, then all storage units in that column that have experienced a CE are classified as Level 2 in terms of risk level, and all storage units in that column that have not experienced a CE are classified as Level 1 in terms of risk level, and that column is marked as a risk column.
[0014] In this example, when the current faulty storage cell experiences this CE, it affects the risk status of other cells in its column. Therefore, marking the risk level makes it easier to accurately determine the risk area later.
[0015] In some possible implementations, determining the risk status of a faulty storage unit based on fault information further includes: if the number of risky rows and / or risky columns in the storage array where the faulty storage unit is located reaches a third threshold within a preset period, then the storage array is marked as a risk layer. Here, the risky behavior is the storage unit where a CE occurs within the preset period and reaches a first threshold; the risky column is the storage unit where a CE occurs within the preset period and reaches a second threshold.
[0016] In this way, the risk status of the entire storage array can be assessed by analyzing the risk rows and columns, which facilitates the accurate identification of risk areas in the future.
[0017] In some possible implementations, determining the risk status of the faulty storage unit and the risk status within the associated storage granules based on fault information also includes: determining the target storage unit based on address information, wherein the target storage unit is a storage unit located in the associated granules and has the same level, row address, and column address as the storage array in which the faulty storage unit is located; and querying the historical risk records of the target storage unit from the fault risk database to obtain the risk status within the associated storage granules.
[0018] In this way, since the risk status of related storage particles in the same memory column affects each other, the risk status of the target storage cell in the related storage particles can be determined by the faulty storage cell, which can realize joint risk assessment across particles and improve the accuracy of risk area assessment.
[0019] In some possible implementations, risk regions of memory are determined based on the risk status of the faulty storage unit and the risk status within the associated storage particles. This includes: if a target storage unit has experienced a Complete Execution (CE) within a preset period: if neither the row nor the column containing the faulty storage unit has experienced a CE within the preset period, then the faulty storage unit is identified as a risk region; if the row containing the faulty storage unit is a risky row, then the row containing the faulty storage unit is identified as a risk region; a risky row is a row in which storage units that have experienced a CE within the preset period reach a first threshold; if the column containing the faulty storage unit is a risky column, then the column containing the faulty storage unit is identified as a risk region; a risky column is a column in which storage units that have experienced a CE within the preset period reach a second threshold; if the storage array containing the faulty storage unit is a risk layer, then the storage array containing the faulty storage unit is identified as a risk region; a risk layer is a storage array in which the number of risky rows or risky columns reaches a third threshold.
[0020] In this example, based on the risk status of the target memory cell across different granules, all risk regions within the memory cell containing the currently faulty memory cell can be determined. Similarly, risk regions in other associated memory cells can also be determined based on the risk status of the faulty memory cell, thus obtaining all risk regions in the entire memory.
[0021] In some possible implementations, risk areas of memory are determined based on the risk status of the faulty storage unit and the risk status within the associated storage particles. This includes: if a target storage unit has experienced a Complete Execution (CE) within a preset period: if neither the row nor the column containing the target storage unit has experienced a CE within the preset period, then the target storage unit is identified as a risk area; if the row containing the target storage unit is a risky row, then the row containing the target storage unit is identified as a risk area; a risky row is a row in which storage units that have experienced a CE within the preset period reach a first threshold; if the column containing the target storage unit is a risky column, then the column containing the target storage unit is identified as a risk area; a risky column is a column in which storage units that have experienced a CE within the preset period reach a second threshold; if the storage array containing the target storage unit is a risk layer, then the storage array containing the target storage unit is identified as a risk area; a risk layer is a storage array in which the number of risky rows or risky columns reaches a third threshold.
[0022] In a second aspect, embodiments of this application provide a computing device including a baseboard management controller, the baseboard management controller being configured to perform the method as described in any possible implementation of the first aspect.
[0023] Thirdly, embodiments of this application provide a management platform, comprising: an acquisition module, configured to acquire fault information, the fault information including the address information and fault time of a faulty storage unit, the faulty storage unit being a storage unit in memory where a correctable error (CE) has occurred, wherein the memory includes at least one memory column, each memory column including multiple memory particles, and each memory particle including several storage units; a processing module, configured to determine the risk status of the faulty storage unit and the risk status within associated storage particles based on the fault information, wherein associated storage particles are memory particles located in the same memory column as the memory particle to which the faulty storage unit belongs, wherein the risk status of the faulty storage unit includes the risk level of the faulty storage unit and the risk level of the memory region surrounding the faulty storage unit, the risk level being used to represent the degree of risk of an uncorrectable error (UCE) occurring, and the risk status within associated storage particles including the risk level of their own internal storage units; the processing module is further configured to determine a risk area of memory based on the risk status of the faulty storage unit and the risk area within associated storage particles, the risk area of memory being an area that needs to be isolated.
[0024] It is understood that the management platform can run on a general-purpose server or on cloud infrastructure, or it can be a software program running on the baseboard management controller (BMC) of a computing device; this embodiment does not limit it to a single instance.
[0025] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.
[0026] Fifthly, embodiments of this application provide a computer program product, characterized in that, when the computer program product is run on a processor, it causes the processor to execute the method described in the first aspect or any possible implementation of the first aspect.
[0027] In a sixth aspect, embodiments of this application provide a chip, characterized in that it includes at least one processor and an interface; the at least one processor obtains program instructions or data through the interface; the at least one processor is used to execute program line instructions to implement the method described in the first aspect or any possible implementation of the first aspect.
[0028] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application;
[0030] Figure 2 This is a schematic diagram of the memory structure in a specific embodiment of this application;
[0031] Figure 3 This is a schematic diagram of a data read operation performed inside a computing device in a specific embodiment of this application;
[0032] Figure 4 This is a schematic diagram illustrating the risk level marking of storage units in a specific embodiment of this application;
[0033] Figure 5 This is a schematic diagram illustrating the determination of risky regions in memory in a specific embodiment of this application;
[0034] Figure 6 This is a schematic diagram of a computing device performing memory fault handling in a specific embodiment of this application;
[0035] Figure 7 This is a schematic flowchart of a memory fault handling method provided in an embodiment of this application;
[0036] Figure 8 This is a flowchart illustrating a memory fault handling method in a specific embodiment of this application;
[0037] Figure 9This is a schematic diagram illustrating a data read operation in a specific example of this application;
[0038] Figure 10 This is a schematic diagram of the structure of a management platform provided in an embodiment of this application;
[0039] Figure 11 This is a schematic diagram of the structure of a chip provided in an embodiment of this application. Detailed Implementation
[0040] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.
[0041] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.
[0042] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.
[0043] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.
[0044] To facilitate understanding of the technical solutions of the embodiments of this application, the terms used herein are explained below.
[0045] CE (corrected error) indicates that an error has been corrected.
[0046] UCE (uncorrected error) indicates an uncorrectable error.
[0047] RAS (reliability, availability, and serviceability) is an abbreviation for reliability, availability, and serviceability.
[0048] DDR (double data rate SDRAM) is short for Double Data Rate Synchronous Dynamic Random Access Memory, which is a type of memory. It can include DDR4, DDR5, and other types, but is not limited to these.
[0049] DIMM (dual-inline-memory-modules) is a type of memory module. For example, the memory chips on a DIMM memory module can be DDR type memory.
[0050] ECC (Error Checking and Correction) is an error correction algorithm used to detect and correct data in memory to ensure data transmission consistency. It is a widely used memory RAS (Recovery and Assurance) function. For example, the ECC mechanism in DDR5 memory chips is an on-die ECC mechanism, which allows the calculation and error correction process to be completed entirely autonomously by the chip, and can correct single-bit errors.
[0051] ADDDC (adaptive double-device data correction) is a technology that can correct multi-bit errors in a single memory chip on a DIMM. It can dynamically isolate a single chip when a persistent failure occurs, and can isolate it two or three times in succession depending on the isolation location.
[0052] SDDC (single-device data correction) is a technology that can correct multi-bit errors in a single memory chip on a DIMM and can dynamically isolate the faulty chip online when a single memory chip experiences a persistent failure.
[0053] BIOS (Basic Input Output System) is responsible for setting various parameters of the system hardware. Essentially, it is a set of programs, or code, that is, a set of programs that are fixed on a memory chip on the motherboard of the computer.
[0054] A burst transfer refers to a method of continuously transmitting data between adjacent memory cells in the same row. As long as the starting address and burst length are specified, read / write operations can be automatically performed on the same number of memory cells in sequence.
[0055] DQ is the data signal during memory read / write operations.
[0056] Typically, CPUs equipped with standard DDR memory (such as DDR4 and DDR5) can implement Side-Band ECC error correction technology (such as Chipkill and SDDC). ECC data can be sent to the memory as sideband data, simultaneously writing this data to multiple memory chips on the memory module. If any one of these chips fails, only the corresponding single byte of data is affected. This error correction technology can correct multi-bit errors (CE) on a single memory chip, but when multiple memory chips fail simultaneously, this error becomes uncorrectable, resulting in an uncorrectable error (UCE). Therefore, the risk of memory module failure can be considered the risk of multiple memory chips failing simultaneously.
[0057] The existing memory fault handling methods involve statistically analyzing historical memory error correction (CE) events. When the number of CEs reaches a threshold, the BIOS (Basic Input / Output System) issues a command to trigger the CPU's built-in low-level memory isolation mechanism, locally isolating the faulty area to prevent further read / write operations. However, because a CE on one memory cell can affect surrounding cells, implicitly increasing the fault risk in those cells, a persistent memory error (UCE) might eventually occur in an area where no CE has previously occurred. Therefore, this fault handling method cannot accurately determine the severity and exact location of the memory fault, resulting in poor fault isolation accuracy and a higher likelihood of system crashes.
[0058] Other processing methods can analyze the fault characteristics of memory cells based on their fault identification (CE) information, and use deep learning models to predict the probability of uninterrupted fault identification (UCE) in the memory cells, thereby identifying and isolating high-risk areas with a high probability of UCE. However, this method only predicts the fault risk of memory cells in a single memory chip in isolation, and the accuracy of the risk assessment is relatively low.
[0059] To improve the accuracy of isolating memory faulty regions, this application provides a memory fault handling method. This method primarily involves statistically analyzing the risk status of each storage cell and its associated regions (including the row, column, and storage array) within a memory module. It then jointly determines the risk status of storage cells located at the same position across memory chips within the same memory rank, thereby identifying and isolating regions with high fault risk. Since a fault erroneous (CE) event on a storage cell increases the fault risk of surrounding storage cells, this method effectively improves the accuracy of risk region location and isolation through joint determination of risk status across associated regions of memory chips. This helps reduce the risk of system crashes caused by memory-related fault erroneous (UCE) events on computing devices.
[0060] To facilitate understanding, a computing device provided in an embodiment of this application will be described below with reference to the accompanying drawings.
[0061] For example, Figure 1 The diagram shown is a structural schematic of a computing device provided in an embodiment of this application. The computing device 100 can be a server, a computer, or a device deployed as a virtualization (such as a container or virtual machine), but is not limited thereto. The server can be a physical server or a cloud server, etc. Figure 1 As shown, the computing device 100 may include a processor 110, a memory 120, a memory module 130, and a communication interface 140. The memory 120, memory module 130, and communication interface 140 can be communicatively connected to the processor 110 via a bus 160. Furthermore, the computing device 100 may also include a Baseboard Management Controller (BMC) 150, which can be connected to the processor 110, memory 120, memory module 130, and other hardware via the bus 160.
[0062] In this example, processor 110 (or central processing unit, CPU) is the computing and control core of computing device 100. It can be used to process access requests from outside computing device 100, as well as to process events and requests generated inside computing device 100, such as processing CE events of memory 130.
[0063] In this example, memory 120 can provide storage resources for computing device 100 to store computer program instructions, including computer operating system (OS) 101, firmware (such as BIOS 102), and various programs. Memory 120 can be a disk or other type of storage medium, such as a solid-state drive or shingled magnetic recording hard drive, or it can be universal flash storage (UFS) or read-only memory (ROM), or any other computer-readable storage medium capable of carrying or storing program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto.
[0064] Memory 130 refers to internal memory that directly exchanges data with the processor. It can read and write data at any time, serving as temporary data storage for OS 101 or other running programs. Memory 130 can be random access memory, such as dynamic random access memory (DRAM), storage class memory (SCM), or static random access memory (SRAM). Memory 130 can also be a dual in-line memory module (DIMM), i.e., a module composed of DRAM. In practical applications, computing device 100 may include multiple memory modules 130, and these memory modules 130 can be of different types.
[0065] As a concrete example, a memory module 130 is a single memory stick entity. A memory module 130 can include two memory ranks 131, which can be denoted as rank0 and rank1 respectively. Each memory rank 131 can include multiple memory devices 132. For example... Figure 2 As shown in (2a), rank1 includes eight memory chips 132, which can be labeled dev_0, dev_1, ..., dev_7. In this example, each memory chip 132 may include multiple storage arrays (banks) 1320, for example... Figure 2 As shown in (2b), taking dev_0 as an example, it includes multiple storage arrays 1320 such as bank0, ..., bankX (X≥1).
[0066] Storage array 1320 is a data storage entity. Each storage array 1320 may include a number of storage cells 1321. The number of storage cells 1321 is determined by the number of rows and columns in the storage array 1320, such as... Figure 2 As shown in (2c). According to the data reading mechanism, each time data is read from memory 130, a portion of the data is read from the same location (meaning that the storage array is at the same level, the row address is the same, and the column address is also the same) of the storage unit 1321 in different memory chips 132 of memory 130, so as to obtain the final target data to be read.
[0067] For example. Figure 3As shown, during a read operation, data is read from the same cell at each memory chip 132 on rank 1 of memory 130. For example, partial data X10, X11, ..., X17 are read from eight cells: cell_0 in the second row and second column of bank 7 in dev_0, cell_1 in the second row and second column of bank 7 in dev_1, ..., and cell_7 in the second row and second column of bank 7 in dev_7. Then, memory 130 transmits all of these data X10, X11, ..., X17 to processor 110 for Side-Band ECC error correction, finally concatenating them to obtain the target data XF for this read operation. In this example, if there is an error in the data from a certain chip in the target data XF (such as data X11 in dev_1), whether it is a single-bit or multi-bit error, it can be corrected based on the Side-Band ECC mechanism, and a corresponding CE event is generated and transmitted to BMC150 for recording.
[0068] In this embodiment, the BMC150 can perform a series of monitoring and control functions on the computing device 100, operating on the system hardware. For example, it monitors the system temperature, voltage, etc., and performs corresponding adjustments to ensure the system is in a healthy state. It is also responsible for recording various hardware information and log entries for monitoring and alarm purposes. The BMC is an independent system that can communicate with other hardware on the computing device 100 (such as the processor 110, memory 130, etc.) through physical channels, and can also interact with the OS 101 and BIOS 102.
[0069] Furthermore, the BMC150 maintains a fault risk database 151, which records the risk status of all storage units 1321 in all memory 130s of the computing device 100. The risk status can include the number and duration of Complete Errors (CEs) in each storage unit 1321, as well as the risk level of the storage unit 1321 itself. For example, for any storage unit 1321, the risk level is related to the number and duration of its own CEs, as well as the number and duration of CEs and their risk levels in other storage units in its row, column, and / or bank. The more CEs a storage unit 1321 experiences within a predetermined period, the higher its risk level; similarly, the more CEs occur in the surrounding area within that period, the higher the risk level. For example, the preset period can be 24 hours, but it is not limited to this. It is understood that a preset period that is too short will affect overhead, while a period that is too long will affect the accuracy of risk assessment.
[0070] The following is based on Figure 4 The example shown illustrates the principle of risk level marking for storage unit 1321. Figure 4Taking a memory cell_01 in memory chip dev_0 as an example, when BMC150 receives a CE event about memory cell_01, it performs the following marking strategy:
[0071] like Figure 4 As shown in (4a), if, within a preset period (e.g., within 24 hours), only storage cell_01 in the row and column where the storage cell_01 is located has experienced this CE, then the risk level of storage cell_01 can be marked as Level 1 (or Medium Risk Level, see [reference]). Figure 4 (4a) Gray-filled area); If, in addition to this CE, the memory cell_01 has experienced other CEs within the preset period, then the risk level of the memory cell_01 is marked as Level 2 (or High Risk Level, see [reference]). Figure 4 (4b) with black filling area), it can be understood that the risk level of the second level is higher than that of the first level, and the storage cell 1321 in the first level or the second level is a risk cell;
[0072] like Figure 4 As shown in (4c), if, besides this CE, another storage cell (such as cell_02) in row 2 of the same row as cell_01 has experienced a CE within the preset period, then the risk level of all cells in row 2 that have experienced a CE is marked as Level 2 (see [reference]). Figure 4 In (4c), the black-filled area), and all cells in row 2 that have not experienced a CE are marked as Level 1 in risk level (see [reference]). Figure 4 In the gray-filled area of (4c), if the storage cell of row2 that has CE within the preset period reaches the first threshold (the first threshold can be set to ≥2, but is not limited to this), it can be regarded as a risky row;
[0073] like Figure 4 As shown in (4d), if, besides this CE, another storage cell (such as cell_03) in column 2, which contains cell_01, has experienced a CE within the preset period, then the risk level of all cells in column 2 that have experienced a CE is marked as Level 2 (see [reference]). Figure 4 (in the black-filled area of (4d)) and all cells in column 2 that have not experienced CE are marked as Level 1 in risk level (see Figure 4 In the gray-filled area of (4d), if the storage cell of column2 that has CE within the preset period reaches the second threshold (the second threshold can be set to ≥2, but is not limited to this), it can be regarded as a risk column;
[0074] like Figure 4 As shown in (4e), if after this CE occurs, within a preset period, the number of risky rows and / or risky columns in the storage array bankX where the storage cell_01 is located reaches a preset threshold (i.e., the third threshold), for example, the third threshold for both the number of risky rows and the number of risky columns is 2, then in (4e), the risky rows include rows 2 and 4, and the risky columns include columns 2 and 4. Since both the number of risky rows and the number of risky columns reach their respective third thresholds, the risk level of all cells in the storage array bankX that have experienced a CE is marked as Level 2 (see [reference]). Figure 4 (4e) Black-filled area), and all cells in the storage array bankX that have not experienced CE are marked as Level 1 in risk level (see Figure 4 In the gray-filled area of (4e), the storage array bankX can be considered a risk layer. It is understandable that the preset thresholds for the number of risk rows and the number of risk columns can also be different.
[0075] Thus, in this embodiment, the risk status of all storage units 1321 in memory 130 is recorded in the fault risk database 151 using the aforementioned marking strategy, and the risk level (i.e., risk degree) of each storage unit 1321 is marked. Furthermore, since ECC error correction operations under the same rank in memory 130 are correlated, in this embodiment, BMC 150 also performs joint risk degree determination across memory particles under the same rank based on the records in the fault risk database 151, so as to more accurately find memory regions with higher risk degrees for isolation.
[0076] For example, within the same memory column 131 of the same memory 130, for any risk cell of any memory chip 132, if there are other memory cells 1321 at the same location that are also risk cells, then the memory 130 is considered to have a fault risk, and the risk area needs to be isolated. The location and extent of the isolation are determined based on the risk range formed by the memory cells 1321. Specifically, refer to... Figure 5 As shown, taking two memory chips dev_0 and dev_1 located in the same rank as an example, if storage cell_01 in dev_0 and cell_11 in dev_1 are storage cells located in the same position, and both cell_01 and cell_11 are risk cells (level 1 or level 2), then the BMC150 will process cell_01 using the following region determination strategy:
[0077] refer to Figure 5As shown in (5a), if the storage cell_01 is an independent risk cell, that is, there are no other risk cells in its row, column, storage array, etc., then it is determined that only cell_01 will be isolated;
[0078] refer to Figure 5 As shown in (5b), if storage cell_01 is located on a risky row2, then it is determined that the risky row2 should be isolated; similarly, refer to Figure 5 As shown in (5c), if the storage cell_01 is located on a risk column2, then it is determined that the risk column2 should be isolated.
[0079] refer to Figure 5 As shown in (5d), if the storage cell cell_01 is located in a risk layer bankX, then the risk layer bankX where the storage cell cell_01 is located is isolated.
[0080] It is understandable that the isolation determination logic for cell_11 is the same as that for cell_01. Furthermore, when there are three or more memory chips 132 with storage cells 1321 at the same location that are considered risk cells, the isolation determination logic for these risk cells is the same as that for cell_01.
[0081] In this way, the BMC150 performs a joint risk assessment of the memory 130 at the cell level across memory particles based on the fault risk database 151, which can locate memory regions with high fault risk for isolation. Because this assessment method integrates the risks of storage cells 1321 at the same location and related regions (such as the row, column, and storage array) between different memory particles 132, it can greatly improve the accuracy of risk area isolation, thereby more effectively reducing the risk of downtime caused by UCE.
[0082] In one possible implementation, such as Figure 6As shown, when a read operation is performed on the computing device 100, the communication process between components may specifically include: the memory 130 transmits the data read from each memory chip 132 to the processor 110 through step 1; the processor 110 can perform ECC error correction processing in step 2 based on the internal hardware error detection and reporting mechanism (machine check architecture, MCA); if a CE event is generated during ECC error correction, information representing the CE event can be recorded in the relevant registers on the processor 110 (such as the CSR (control and status register) register and the MSR (model-specific register) register); then the processor 110 reports the CE event to the BIOS 102 through step 3 and transmits the relevant information recorded in these registers to the BIOS 102; and the BIOS 102 transmits the information to the BMC 150 through step 4. Next, step 6 is executed in BMC150. Based on the CE event, the aforementioned marking and region determination strategies are applied to memory 130 in the fault risk database 151. If the processing result indicates that risk region isolation of memory 130 is not required, only the fault risk database 151 needs to be updated. If the processing result indicates that certain risk regions need to be isolated, BMC150 can report the information of these risk regions to OS101 through step 7, notifying OS101 to perform corresponding soft isolation (such as pageoff processing), or it can instruct the CPU to perform corresponding hard isolation (such as replacing the risk region using redundant space) through step 8. This completes the isolation operation for the risk regions of memory 130.
[0083] In some other possible implementations, the aforementioned fault risk database 151 can also be maintained by the processor 110. Accordingly, the aforementioned marking strategy and region determination strategy executed by the BMC 150 are also executed by the processor 110, which will not be elaborated here.
[0084] In some other possible implementations, the maintenance of the aforementioned fault risk database 151, as well as the execution of the marking strategy and the region determination strategy, can also be implemented on an external computing device or cloud management platform of the computing device 100.
[0085] In this embodiment, the computing device 100 realizes communication between various modules, devices, units, components and / or equipment through the communication interface 140.
[0086] In this embodiment, bus 160 includes a bus using one or more communication protocols. Furthermore, bus 160 includes hardware, software, or both, that couples components of computing device 100 together. For example, and not limitingly, bus 160 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI E) bus, a Serial Advanced Technology Attachment (SATA) bus, or other suitable buses, or combinations of two or more of these. It should be understood that although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.
[0087] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the computing device 100. In other embodiments of this application, the computing device 100 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.
[0088] Next, based on the content described above, a memory fault handling method provided by an embodiment of this application will be introduced. It is understood that this method is proposed based on the content described above, and some or all of the content of this method can be found in the description above.
[0089] Please see Figure 7 , Figure 7 This is a flowchart illustrating a memory fault handling method provided in an embodiment of this application. It is understood that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities, for example in... Figure 1 The execution is performed on the BMC150 of the computing device 100 shown, or on the BMC, as will be used hereinafter. Figure 7 As shown, the method may include:
[0090] In S710, fault information is obtained, including the address information of the fault storage unit and the fault time. The fault storage unit is the storage unit in memory where a correctable error (CE) occurred.
[0091] In this embodiment, when the memory 130 on the computing device 100 performs a read operation, it transmits all the data read from multiple memory chips 132 to the processor 110 for ECC error correction. If an error occurs in the data of a storage cell 1321 (i.e., a faulty storage cell) in a certain memory chip 132 (such as a single-bit error or a multi-bit error), then the faulty storage cell will experience an ECC error.
[0092] In some possible examples, after the processor 110 determines that a faulty memory cell has experienced a Complete Execution (CE), it can transmit fault information such as the address information and fault time of the faulty memory cell to the BMC 150 via the BIOS 102. In this way, the BMC 150 can execute step S710 to determine the appropriate risk level of the memory 130.
[0093] For example, the address information is used to describe the identifier of the memory 130 where the faulty storage unit is located, the identifier of the memory column 131 where it is located, the identifier of the memory chip 132 where it is located, the level of the storage array, and the row and column addresses, etc. Each identifier can be a number or a name, etc. The fault time is the time when the faulty storage unit experiences a CE (Collision Error).
[0094] S720 determines the risk status of the faulty storage unit and the risk status within the associated storage particles based on the fault information.
[0095] In this embodiment, the BMC150 can determine the risk status of the faulty storage unit by querying the historical risk records of the faulty storage unit and / or the historical risk records of the surrounding storage units (i.e., the surrounding memory area) within a preset period in the fault risk database 151 based on the fault information maintained internally. The risk status of the faulty storage unit can include the risk level of the faulty storage unit and the risk levels of its surrounding storage units. The risk level can include multiple levels (such as no risk, medium risk, or high risk, but is not limited to these) to respectively represent the degree of risk of UCE occurring.
[0096] For example, historical risk records may include the number of CE occurrences, the time of occurrence, and the risk level of the corresponding storage unit.
[0097] For example, if, based on historical risk records, the faulty storage unit and its surrounding memory area (such as its row, column, and storage array area) have not experienced any Complete Execution (CE) events within a preset period, the frequency of CE events occurring in the faulty storage unit can be considered low, and therefore its risk level is low. Its risk level can be marked as Level 1 (or Medium Risk Level). The preset period can be customized, for example, set to 24 hours, but is not limited to this.
[0098] In this example, if no CE (Collision Error) has occurred in the memory area surrounding the faulty memory cell within a preset period, but the faulty memory cell itself has experienced CE in the past, it can be considered that the frequency of CE occurrences in the faulty memory cell is relatively high, and therefore the risk level is high. Its risk level can be marked as Level 2 (or High Risk). Similarly, if no CE has occurred in the faulty memory cell within a preset period, but the number of CE occurrences in the surrounding memory area and / or the number of memory cells experiencing CE in the past reach preset risk conditions (e.g., the number of CE occurrences reaches a set value, the number of cells experiencing CE reaches a set value, etc.), it can be considered that the faulty memory cell is easily affected by the surrounding memory cells and therefore has a high risk level. Its risk level can be marked as Level 2 (or High Risk).
[0099] Similarly, in this example, the memory cells around the faulty memory cell will also be affected by the CE (Critical Error) that occurs in the faulty memory cell. If the number of CEs in the surrounding memory area and / or the number of memory cells that have CEs after the faulty memory cell has CEs reach the preset risk conditions, then the risk level of all memory cells (including the faulty memory cell) in the surrounding memory area can be improved, for example, from no risk to level one, from level one to level two, and so on.
[0100] For example, the risk status of a faulty storage unit determined based on fault information is recorded in the fault risk database 151.
[0101] The S730 determines the risk areas of memory based on the risk status of the faulty memory cell and the risk status within the associated memory chip, and then isolates the risk areas.
[0102] In this embodiment, the risk situation within associated storage particles is also determined based on the address information of the faulty storage unit, where associated storage particles are memory particles 132 located in the same memory column 131 as the memory particle to which the faulty storage unit belongs. Since data is read from storage particles 1321 at the same location (i.e., the same row and column addresses in the same level of storage arrays across multiple memory particles in the same memory column) in the same memory column, the risk level of the faulty storage unit and the risk situation of storage particles at the same location in its associated storage particles can both serve as reference factors for the potential UCE risk on memory 130, and these reference factors are correlated. If most of the storage particles (including the faulty storage unit) at the same location in different memory particles within the same memory column 131 are at medium or high risk, the risk of UCE will be significantly increased. Therefore, in this embodiment, the risk area can also be determined by querying the risk situation within the associated storage particles of the faulty storage unit.
[0103] Specifically, in this embodiment, after BMC150 updates the fault risk database 151 based on the fault information, it can determine the risk area based on the risk status of the faulty storage unit and the risk status within the associated storage particles. The risk status within the associated storage particles mainly refers to the risk status of storage units in the same bank at the same level as the faulty storage unit, including the risk status of cells located in the same position as the faulty storage unit, the risk status of other cells in the same row / column, and the risk status of the entire bank, etc.
[0104] If both the faulty storage cell and its associated storage particles are at medium or high risk, these medium and high risk areas can be designated as risk areas. In addition, if both the faulty storage cell and its associated storage particles are at medium or high risk, and the number of historical CE occurrences or the number of storage cells that have experienced CE in the row, column, or storage array where the faulty storage cell is located reaches a preset threshold, these rows, columns, and storage arrays can also be designated as risk areas.
[0105] Next, the risk area can be isolated by using the underlying isolation logic built into the processor 110 or OS 101 of the computing device 100, such as redundant space replacement or memory page offline, which will not be listed here.
[0106] Thus, in this embodiment, the correlation between the risk status of the faulty storage unit and its surrounding memory area, as well as the correlation between the risk status of the faulty storage unit and its associated storage particles, are used as the basis for judgment. This allows for joint judgment of risk areas across memory particles, improving the accuracy of risk area isolation and greatly reducing the probability of UCE.
[0107] The memory fault handling method in a specific embodiment of this application will be described in detail below.
[0108] For example, Figure 8 The diagram shows a flowchart of a memory fault handling method according to a specific embodiment of this application.
[0109] like Figure 8 As shown, the method may specifically include:
[0110] S810, the processor of the computing device generates a CE event for target data, which is data read from memory.
[0111] In this step, when the processor 110 of the computing device 100 performs a read operation on the memory 130, it reads partial data from storage cells 1321 at the same location among the multiple memory chips 132 in the memory 130, thereby obtaining the target data. Storage cells 1321 at the same location refer to cells located in different memory chips 132 under the same memory column 131 in the memory 130, and whose respective storage array levels, row addresses, and column addresses are the same.
[0112] For example. (Reference) Figure 9 As shown, during a single read operation, data X10, X11, ..., X17 are read from the same storage unit 1321 in memory chips dev_0, dev_1, ..., dev_7 of memory 130, respectively. This memory 130 is a DDR5 type supporting On-Die ECC, which allows for On-Die ECC error correction (correcting single-bit errors) within the memory chip itself. For example... Figure 9 An error occurs in the data X111 read from memory chip dev_1. This error can be corrected internally to obtain X11a. However, errors in multiple bits (X161 and X162) in memory chip dev_6 cannot be corrected and need to be passed to processor 110 for Side-Band ECC error correction, ultimately obtaining the target data XF. Processor 110 generates a corresponding CE event based on the error occurring in a single chip dev_6. It can be understood that, compared to the On-Die ECC mechanism inside memory chip 132, the Side-Band ECC mechanism is a global ECC error correction mechanism. Errors occurring on memory chip 130 can be recorded in the status register and reported to the relevant registers on processor 110. Memory chip 130 can also be a memory type that does not support the On-Die ECC mechanism, in which case ECC error correction is not performed internally by memory chip 132.
[0113] For example, if the faulty storage unit 1321 is considered a faulty storage unit, then the CE event encapsulates fault information. This fault information may include the address information and the time of the faulty storage unit. The address information describes the identifier of the memory 130 where the faulty storage unit is located, the identifier of the memory column 131, the identifier of the memory chip 132, the level of the storage array, and the row and column addresses, etc. Each identifier may be a number or a name, etc. The time of the fault is the time when the faulty storage unit experiences a CE.
[0114] In the S820, the processor transmits CE events to the BIOS.
[0115] In this step, the CE event generated by the processor 110 is reported to the BIOS 102, and the BIOS 102 executes step S821 to pass the CE event to the BMC 150.
[0116] S830, BMC obtains fault information from CE events.
[0117] S840 determines the risk status of the faulty storage unit and the risk status within the associated storage particles based on the fault information.
[0118] In this example, BMC150, through step S830, parses the address information and failure time of the faulty storage unit from the CE event, and then determines the risk status of the faulty storage unit based on the fault risk database 151. The risk status of the faulty storage unit can include the number of CEs occurring within a preset period, the time of each CE, and the risk level. The risk level is related to the number and time of CEs occurring in the faulty storage unit itself, as well as the number and time of CEs occurring in other storage units in its row, column, and / or bank, and their risk levels. Furthermore, in this example, BMC150 also updates the risk status of the surrounding memory region based on this CE event. This surrounding memory region can include the area formed by the row, column, and storage array of the faulty storage unit.
[0119] As a specific example, the execution of step S840 may include the following steps S841 to S843:
[0120] S841, based on the fault information, determines the risk status of the faulty storage unit.
[0121] In this example, please refer again. Figure 4 As shown in (4a), if within a preset period, the faulty storage unit (in this example, i.e., Figure 4 If a single CE occurs only once in the row and column of cell_01, the risk level of the faulty storage unit can be determined as Level 1 (medium risk), and the time, number of times, and risk level of the CE occurrence for that faulty storage unit can be recorded in the fault risk database 151.
[0122] In this specific example, if, within a preset period, besides this CE, the faulty memory cell has experienced other CEs in the past, and only this faulty memory cell has experienced a CE in its row and column, such as... Figure 4 As shown in (4b), the risk level of the faulty storage unit can be determined as Level 2 (high risk).
[0123] In this specific example, such as Figure 4As shown in (4c), if, within a preset period, besides this CE, other storage units in the row containing the faulty storage unit have also experienced CE, then the risk level of all storage units 1321 in that row that have experienced CE (and occurred before this CE) is marked as Level 2, and the risk level of all storage units 1321 in that row that have not experienced CE is marked as Level 1. In this case, that row is a risky row. Similarly, as... Figure 4 As shown in (4d), if, within a preset period, besides this CE, other memory cells in the column containing the faulty memory cell have also experienced CE, then the risk level of all memory cells 1321 in that column that have experienced CE (and occurred before this CE, the same below) is marked as Level 2, and the risk level of all memory cells 1321 in that column that have not experienced CE is marked as Level 1. In this case, that column is a risk column. Similarly, as... Figure 4 As shown in (4e), if after a CE occurs within a preset period, the number of risk rows and / or risk columns in the storage array where the faulty storage unit is located reaches a preset threshold, then the risk level of all storage units 1321 in the storage array that have experienced a CE is marked as the second level, and the risk level of all storage units 1321 in the storage array that have not experienced a CE is marked as the first level. At this time, the storage array is a risk layer.
[0124] In this way, BMC150 can update the risk status of the faulty storage unit and its surrounding memory area based on the historical records of the fault risk database 151 at each CE. This means that the impact of CE on the faulty storage unit and its surrounding memory area can be visualized through this risk level marking method, which serves as the basis for determining the risk area and helps to improve the accuracy of risk area isolation.
[0125] In this step, after BMC150 updates the fault risk database 151 based on the CE event, it can identify risk areas (areas prone to UCE risks) based on the risk status of the faulty storage unit and its surrounding memory area in the fault risk database 151, combined with the risk status of other memory particles 132 (i.e., associated storage particles) under the same memory column 131.
[0126] Therefore, executing step S840 may specifically include:
[0127] S842 determines the target memory unit based on the address information of the faulty memory unit.
[0128] In this example, the target storage cell is a storage cell located in the associated storage cell of the faulty storage cell and at the same location as the faulty storage cell. The associated storage cell is a memory cell located in the same memory column 131 as the memory cell containing the faulty storage cell.
[0129] Specifically, after updating the fault risk database 151 based on the CE event, the BMC150 can search for target storage cells in the same rank (memory column 131 where the faulty storage cell is located) among all associated storage particles that are located at the same location as the faulty storage cell (same storage array level, same row address, and same column address). There can be multiple target storage cells.
[0130] S843 queries the historical risk records of the target storage unit from the fault risk database to obtain the risk situation inside the associated storage particle.
[0131] In this example, BMC150 can query the historical risk records of the target storage cell in the fault risk database 151, and can also query the historical risk records of the surrounding memory area of the target storage cell to summarize the risk situation inside the associated storage particle.
[0132] The S850 determines the risk areas of memory based on the risk status of the faulty memory cell and the risk status within the associated memory chip.
[0133] In this example, the BMC150 jointly assesses the risk of the faulty memory cell, each target memory cell, and the memory regions surrounding the faulty memory cell and the target memory cells.
[0134] For example, you can refer to [the example again]. Figure 5 As shown, the faulty storage unit (in this example, i.e.) Figure 5 If the risk level of cell_01 shown is either Level 1 or Level 2, and the target storage cell in the associated storage particle 132 of the faulty storage cell is also a risky cell, then BMC150 can process the faulty storage cell and its surrounding memory area according to the preset area determination strategy as follows:
[0135] If a faulty storage cell is an independent risk cell in the storage array, then the faulty storage cell is determined to be a risk area.
[0136] If the row containing the faulty storage unit is a risky row, then that risky row is determined to be a risky area; similarly, if the column containing the faulty storage unit is a risky column, then that risky column is determined to be a risky area; if the storage array containing the faulty storage unit is a risky layer, then the entire risky layer is determined to be a risky area.
[0137] Similarly, the above-mentioned region determination strategy is also applied to each target storage unit and its surrounding memory area, thereby summarizing all the identified risk areas to obtain the risk areas of memory 130.
[0138] It is understandable that, for each associated storage particle of the faulty storage cell in the same memory column 131, if any of the associated storage particles has no risk in the cell located at the same position as the faulty storage cell, then the associated storage particle is not at risk, and there is no need to determine the risk area of the associated storage particle in step S850.
[0139] Thus, this embodiment performs joint determination across memory chips based on the risk status of the faulty memory cell and the target memory cell, thereby determining whether each CE event leads to a high-risk area and the scope of the risk area.
[0140] In the S860, the BMC transmits the address information of the risky region to the OS and / or processor to trigger the isolation operation of the risky region.
[0141] In this step, BMC150 aggregates the address information of all risk regions and triggers the underlying memory isolation action built into processor 110 or OS101 based on the attributes of the risk regions to perform local isolation of the risk regions.
[0142] For example, if the risk area is an independent storage unit, a separate row, or a column, BMC150 can, through step S861, invoke the soft isolation operation of OS101 (such as pageoffline) to take the memory page belonging to the risk area offline in the page table maintained by OS101, making it unavailable and preventing further read and write operations in this area. If the risk area is a storage array, BMC150 can, through step S862, invoke the hard isolation operation of processor 110 (such as ADDDC, but not limited to this) to replace the risk area with redundant space, migrating the data or threads originally stored in the risk area to the designated redundant space, and the risk area will no longer be used.
[0143] Thus, this embodiment, based on the principles of CE and UCE occurrence (i.e., ECC error correction mechanism) and the correlation between the faulty storage unit and its surrounding memory area and associated storage particles, can determine whether memory 130 has a fault risk and provide a more accurate isolation location and range, thereby helping to reduce the probability of UCE occurrence.
[0144] In other possible implementations, the fault risk database 151 can be maintained by the CPU, and correspondingly, the steps to be performed by the BMC 150 can also be performed by the processor 110, which will not be elaborated further.
[0145] Based on the methods described in the above embodiments, this application provides a management platform. Please refer to... Figure 10 , Figure 10 This is a schematic diagram of the structure of a management platform provided in an embodiment of this application.
[0146] like Figure 10 As shown, the management platform 1000 may include an acquisition module 1001 and a processing module 1002. The acquisition module 1001 can acquire fault information, including the address information and fault time of the faulty storage unit. The faulty storage unit is a storage unit in memory where a correctable error (CE) has occurred. The memory includes at least one memory column, each memory column includes multiple memory particles, and each memory particle includes several storage units. The processing module 1002 can determine the risk status of the faulty storage unit based on the fault information. The associated storage particle is a memory particle located in the same memory column as the memory particle to which the faulty storage unit belongs. The risk status of the faulty storage unit includes the risk level of the faulty storage unit and the risk level of the memory region surrounding the faulty storage unit. The risk level indicates the degree of risk of an uncorrectable error (UCE). The risk status within the associated storage particle includes the risk level of its own internal storage units. The processing module 1002 can also determine the risk area of the memory based on the risk status of the faulty storage unit and the risk status within the associated storage particle. The risk area of the memory is the area that needs to be isolated.
[0147] It should be understood that the management platform 1000 is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the management platform 1000 are similar to those described in the above methods. The working process of the management platform 1000 can be referred to the corresponding process in the above methods, and will not be repeated here.
[0148] In some possible implementations, the management platform 1000 can be a software program running on a general-purpose server or a cloud server (such as a virtual machine or container), or the management platform 1000 can also run on a BMC150.
[0149] Based on the methods in the above embodiments, this application provides an electronic device. The electronic device may include a BMC, which is used to execute the methods in the above embodiments.
[0150] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.
[0151] Based on the methods in the above embodiments, this application provides a computer program product, characterized in that, when the computer program product is run on a processor, the processor executes the methods in the above embodiments.
[0152] Based on the methods described in the above embodiments, this application also provides a chip. Please refer to... Figure 11 , Figure 11 This is a schematic diagram of a chip structure provided in an embodiment of this application. Figure 11 As shown, chip 1100 includes one or more processors 1101 and interface circuitry 1102. Optionally, chip 1100 may also include a bus 1103. Wherein:
[0153] Processor 1101 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by integrated logic circuits in the hardware of processor 1101 or by instructions in software form. The processor 1101 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods and steps disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor.
[0154] The interface circuit 1102 can be used to send or receive data, instructions or information. The processor 1101 can use the data, instructions or other information received by the interface circuit 1102 to process the data, instructions or other information, and can send the processed information out through the interface circuit 1102.
[0155] Optionally, chip 1100 also includes memory, which may include read-only memory and random access memory, and provides operation instructions and data to the processor. A portion of the memory may also include non-volatile random access memory (NVRAM).
[0156] Optionally, the memory stores executable software modules or data structures, and the processor can execute corresponding operations by calling the operation instructions stored in the memory (which may be stored in the operating system).
[0157] Optionally, the interface circuit 1102 can be used to output the execution results of the processor 1101.
[0158] It should be noted that the functions of the processor 1101 and the interface circuit 1102 can be implemented through hardware design, software design, or a combination of hardware and software; no restrictions are imposed here.
[0159] It should be understood that each step of the above method embodiments can be completed by hardware logic circuits or software instructions in a processor.
[0160] It is understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. In addition, in some possible implementations, each step in the above embodiments may be selectively executed according to the actual situation, and may be partially or fully executed, which is not limited here.
[0161] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.
[0162] The method steps in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.
[0163] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).
[0164] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application.
Claims
1. A memory failure handling method, characterized by, The method includes: Obtain fault information, which includes the address information and fault time of the fault storage unit. The fault storage unit is a storage unit in memory where a correctable error (CE) occurs. The memory includes at least one memory column, each memory column includes multiple memory particles, and each memory particle includes several storage units. Based on the fault information, the risk status of the faulty storage unit and the risk status within the associated storage particle are determined. The associated storage particle is a memory particle located in the same memory column as the memory particle to which the faulty storage unit belongs. The risk status of the faulty storage unit includes the risk level of the faulty storage unit and the risk level of the memory region surrounding the faulty storage unit. The risk level is used to indicate the degree of risk of an uncorrectable error (UCE). The risk status within the associated storage particle includes the risk level of its own internal storage units. Determining the risk status of the faulty storage unit and the risk status within the associated storage particle based on the fault information further includes: determining a target storage unit based on the address information. The target storage unit is a storage unit located within the associated storage particle and has the same level, row address, and column address as the storage array in which the faulty storage unit is located; querying the historical risk records of the target storage unit from the fault risk database to obtain the risk status within the associated storage particle. Based on the risk status of the faulty storage unit and the risk status within the associated storage particles, a risk area for the memory is determined, and this risk area is the region that needs to be isolated. Specifically, determining the risk area for the memory based on the risk status of the faulty storage unit and the risk status within the associated storage particles includes: If a CE (Collision Error) occurs in the target storage unit within a preset period: if neither the row nor the column containing the faulty storage unit has experienced a CE within the preset period, then the faulty storage unit is identified as a risk area; if the row containing the faulty storage unit is a risky row, then the row containing the faulty storage unit is identified as a risk area; the risky row is the row where the number of storage units experiencing CEs within the preset period reaches a first threshold; if the column containing the faulty storage unit is a risky column, then the column containing the faulty storage unit is identified as a risk area; the risky column is the column where the number of storage units experiencing CEs within the preset period reaches a second threshold; if the storage array containing the faulty storage unit is a risk layer, then the storage array containing the faulty storage unit is identified as a risk area; the risk layer is a storage array where the number of risky rows or risky columns reaches a third threshold. If a CE (Collision Error) occurs in the target storage unit within a preset period: if neither the row nor the column containing the target storage unit has experienced a CE within the preset period, the target storage unit is identified as a risk area; if the row containing the target storage unit is a risky row, the row containing the target storage unit is identified as a risk area; the risky row is the row where the number of storage units experiencing CEs within the preset period reaches a first threshold; if the column containing the target storage unit is a risky column, the column containing the target storage unit is identified as a risk area; the risky column is the column where the number of storage units experiencing CEs within the preset period reaches a second threshold; if the storage array containing the target storage unit is a risk layer, the storage array containing the target storage unit is identified as a risk area, the risk layer being a storage array where the number of risky rows or risky columns reaches a third threshold.
2. The method according to claim 1, characterized in that, The step of determining the risk status of the faulty storage unit based on the fault information includes: Based on the fault information, query the historical risk records of the faulty storage unit and the surrounding memory area from the fault risk database. The historical risk records include historical risk levels. Based on the historical risk records of the faulty storage unit and the surrounding memory region, the risk level of the faulty storage unit and the risk level of the surrounding memory region are updated to obtain the risk status of the faulty storage unit.
3. The method of claim 2, wherein, Each memory chip includes a multi-layer storage array, and each layer of the storage array includes a plurality of storage cells arranged in a matrix; The risk level of the faulty storage unit and the risk level of the surrounding memory region are updated to obtain the risk status of the faulty storage unit, including: If, within a preset period, the faulty storage unit has never experienced a CE (Critical Error) in its history, and the number of CEs and / or the number of storage units experiencing CEs in the row, column, and storage array where the faulty storage unit is located do not reach the preset risk conditions, then the risk level of the faulty storage unit is determined to be Level 1. If, within the preset period, the faulty storage unit has historically experienced a CE (Critical Error), and the number of CEs and / or the number of storage units experiencing CEs in the row, column, and storage array where the faulty storage unit is located do not reach the preset risk conditions, then the risk level of the faulty storage unit is determined to be Level 2, and the risk level of Level 2 is higher than that of Level 1.
4. The method according to claim 2 or 3, characterized in that, The step of determining the risk status of the faulty storage unit based on the fault information further includes: If, within a preset period, other storage units in the row containing the faulty storage unit have also experienced a previous CE (Critical Error), then the risk level of all storage units in that row that have experienced a CE is determined to be Level 2, and the risk level of all storage units in that row that have not experienced a CE is determined to be Level 1. The row is then marked as a risky row. Within a preset period, if other storage units in the column containing the faulty storage unit have also experienced a CE (Critical Error) in the past, then the risk level of all storage units in that column that have experienced a CE is determined to be Level 2, and the risk level of all storage units in that column that have not experienced a CE is determined to be Level 1. This column is then marked as a risk column.
5. The method of claim 2, wherein, The step of determining the risk status of the faulty storage unit based on the fault information further includes: If the number of risky rows and / or risky columns in the storage array where the faulty storage unit is located reaches a third threshold within a preset period, the storage array is marked as a risk layer. The risky rows are the rows in which storage units that have experienced a Complete Execution (CE) within the preset period reach a first threshold; the risky columns are the columns in which storage units that have experienced a CE within the preset period reach a second threshold.
6. A computing device, characterized in that, Includes a substrate management controller, the substrate management controller being used to perform the method as described in any one of claims 1-5.
7. A management platform, characterized in that, include: The acquisition module is used to acquire fault information, which includes the address information and fault time of the fault storage unit. The fault storage unit is a storage unit in memory where a correctable error (CE) occurs. The memory includes at least one memory column, each memory column includes multiple memory particles, and each memory particle includes several storage units. The processing module is configured to determine the risk status of the faulty storage unit and the risk status within the associated storage particle based on the fault information. The associated storage particle is a memory particle located in the same memory column as the memory particle to which the faulty storage unit belongs. The risk status of the faulty storage unit includes the risk level of the faulty storage unit and the risk level of the memory region surrounding the faulty storage unit. The risk level is used to indicate the degree of risk of an uncorrectable error (UCE). The risk status within the associated storage particle includes the risk level of its own internal storage units. The process of determining the risk status of the faulty storage unit and the risk status within the associated storage particle based on the fault information further includes: determining a target storage unit based on the address information. The target storage unit is a storage unit located within the associated storage particle and has the same level, row address, and column address as the storage array in which the faulty storage unit is located; and querying the historical risk records of the target storage unit from the fault risk database to obtain the risk status within the associated storage particle. The processing module is further configured to determine a risk region of the memory based on the risk status of the faulty storage unit and the risk status within the associated storage particles, wherein the risk region of the memory is an area that needs to be isolated; wherein, determining the risk region of the memory based on the risk status of the faulty storage unit and the risk status within the associated storage particles includes: If a CE (Collision Error) occurs in the target storage unit within a preset period: if neither the row nor the column containing the faulty storage unit has experienced a CE within the preset period, then the faulty storage unit is identified as a risk area; if the row containing the faulty storage unit is a risky row, then the row containing the faulty storage unit is identified as a risk area; the risky row is the row where the number of storage units experiencing CEs within the preset period reaches a first threshold; if the column containing the faulty storage unit is a risky column, then the column containing the faulty storage unit is identified as a risk area; the risky column is the column where the number of storage units experiencing CEs within the preset period reaches a second threshold; if the storage array containing the faulty storage unit is a risk layer, then the storage array containing the faulty storage unit is identified as a risk area; the risk layer is a storage array where the number of risky rows or risky columns reaches a third threshold. If a CE (Collision Error) occurs in the target storage unit within a preset period: if neither the row nor the column containing the target storage unit has experienced a CE within the preset period, the target storage unit is identified as a risk area; if the row containing the target storage unit is a risky row, the row containing the target storage unit is identified as a risk area; the risky row is the row where the number of storage units experiencing CEs within the preset period reaches a first threshold; if the column containing the target storage unit is a risky column, the column containing the target storage unit is identified as a risk area; the risky column is the column where the number of storage units experiencing CEs within the preset period reaches a second threshold; if the storage array containing the target storage unit is a risk layer, the storage array containing the target storage unit is identified as a risk area, the risk layer being a storage array where the number of risky rows or risky columns reaches a third threshold.
Citation Information
Patent Citations
Memory management method and device
CN106021014A
Fault processing method and computing device
CN114968652A
