TWI670721B - Unusual power-off test method and device for storage device - Google Patents

Abstract

The invention discloses an abnormal power-off test device for a storage device, comprising a system control unit, a workload simulation unit, a power control unit, a protocol control unit and a data temporary storage unit. The power control unit and the protocol control unit selectively connect the storage device. The test device simulates at least one workload at the storage device and powers off at a specified time and then tests. In this way, the test device quickly discovers potential problems and algorithmic vulnerabilities of the device firmware of the storage device, and improves the stability and reliability of the device firmware of the storage device.

Description

Unusual power-off test method and device for storage device

The invention belongs to a power-off test technology of a storage device, in particular to an abnormal power-off test method and device for a storage device, and the potential of the device firmware can be discovered more quickly through the design of the programmable test. Problems and algorithmic loopholes to improve the stability and reliability of the storage device.

Most solid-state disks (SSDs), which are storage devices, are connected to the busbars of a computer using SATA, SAS, or Fibre Channel. But SATA is mainly used as a hard disk drive (HDD) port, and it is increasingly difficult to meet the ever-increasing speed of SSD over time. Therefore, some SSDs have been limited by the maximum throughput of SATA. Non-volatile memory express (NVMe) is a logical device specification, which is an extensible host controller interface with optimization and high performance. The basic concept of the NVMe is through fast A peripheral component interconnect express (PCIe) bus is connected to the transmission loop of the processor. Its biggest advantage is that it provides ultra-low latency and high-bandwidth access performance, and NVMe adds automatic power state switching and dynamic energy management, which makes NVMe have a big advantage in power management.

When the storage device is shipped from the factory, it must be tested for different levels of durability and workload (such as continuous reading and writing, random reading and writing, reading and writing large files or small files, etc.) according to consumer or enterprise level. Therefore, its test requirements and test methods are different. Standard specifications (such as JESD 218, JESD 219, etc.) to ensure the durability and quality of the storage device. Most of the existing storage device testing methods are repeated tests for specific workloads or increase the system execution time, and refer to the standard specifications to understand the test requirements and test environment. The durability and reliability of the storage device can often be verified by past experience or by lengthening the test time. The important thing about storage devices is the preservation of user data. Device firmware of a typical storage device may handle normal power outages, but it is easy to ignore or not fully consider the abnormal power outage to create a loophole. If the storage device just happens to encounter an abnormal power failure that is not considered, the user's data disappears or the system collapses, and the device firmware of the storage device fails to access the data.

An application is typically executed by the operating system to test existing storage devices. Therefore, the flexibility and complexity of the test process is limited. In addition, such tests are typically periodic power outages that fail to perform predictable abnormal power outage tests. Alternatively, the test device can be powered off randomly by the tester, but the power-off test cannot be performed at the specified time in the different workloads of the storage device, and it is difficult to find the problem of the device firmware when the power is abnormally interrupted under different workloads. And vulnerabilities. In addition, the existing test equipment can only be executed in the same test mode such as the instruction power-off test, the process power-off test or the stress test in each test, and the tester cannot be changed at any time. How to solve the above problems is an important issue in the industry and is also an inventor of the present invention.

The reason is that the present invention is based on the above-mentioned needs and problems, and actively seeks solutions through the experience of the inventors for years of relevant development, and has continuously developed a kind of storage for continuous research and development. The abnormal power-off test method of the device and its abnormal power-off test equipment can more effectively find out the problems and loopholes of the device firmware.

Therefore, the main object of the present invention is to provide an abnormal power-off test method for a storage device and an abnormal power-off test device thereof, so that the test can be performed for the simulation to perform the test under abnormal power failure, to further Quickly discover potential problems with device firmware and algorithmic loopholes, and improve the stability and reliability of device firmware for storage devices.

Another object of the present invention is to provide an abnormal power-off test method for a storage device and an abnormal power-off test device thereof, which can perform a power-off test at a specified point in a different workload of the storage device. In the process of electricity, the issued command does not need to wait for the response of the command, but performs power-off at any time after the command is sent, and performs a valid test through a sudden power-off.

Another object of the present invention is to provide an abnormal power-off test method for a storage device and an abnormal power-off test device thereof, which can change different test modes, test times and test cycles at any time according to requirements to increase storage. The variety of device error tests, Error Cases, Situations, Samples, or Categories.

To achieve the above object, the test device simulates at least one workload in at least one storage device according to at least one test specification, and performs at least one test at a specified time. The test equipment includes a test equipment including a system control unit, a work load simulation unit, a power control unit, a protocol control unit, and a data temporary storage unit. The system control unit can calculate, accept and reply the instructions transmitted during the test. The workload simulation unit is connected to the system control unit, and can execute various workload simulations to be tested on the storage device according to the instruction. The power control unit is connected to the system control unit, and can selectively connect the storage device, and performs power-off and power-on of the storage device according to the instruction. The protocol control unit is coupled to the system control unit and is selectively connectable to the storage device and sends instructions related to the test protocol. The data temporary storage unit is connected to the system control Unit, and can record key information and information to check the correctness of the data stored in the storage device. The test device is programmable.

The following is a description of the preferred embodiments of the present invention, and the detailed description of the present invention will be described in detail with reference to the accompanying drawings. .

10‧‧‧Test equipment

11‧‧‧System Control Unit

12‧‧‧Workload Simulation Unit

13‧‧‧Power Control Unit

14‧‧‧Protocol Control Unit

15‧‧‧Data temporary storage unit

20‧‧‧Storage device

21‧‧‧Device Control Unit

22‧‧‧Flash memory

The first figure is the abnormal power-off test device for the storage device of the present invention. Schematic diagram of the architecture to illustrate its main configuration state.

Second: is a flow chart of the method for testing an abnormal power failure of the storage device of the present invention.

Third Figure: Flowchart of a portion of the method illustrated in the second figure.

As shown in the first figure, an abnormal power-off test device 10 for a storage device 20 includes a system controller unit 11 and a workload load unit. 12. A power control unit 13, a protocol control unit 14, and a data buffer unit 15. The test device 10 is selectively connected to the storage device 20 by the power control unit 13 and the protocol control unit 14 for executing the storage device 20 for different workloads (including but not limited to continuous reading and writing, random reading and writing, reading and writing large files or small Unusual power outage test under file, etc.). The storage device 20 can be an NVMe solid state drive and has a series of Nand Flash 22 controlled by an NVMe Device Controller Unit 21 for programmability testing, allowing the test device 10 to simulate Performing a test on the storage device 20 under abnormal power failure, The potential problems and algorithmic vulnerabilities of the device firmware of the storage device 20 are quickly discovered, and the stability and reliability of the device firmware of the storage device 20 are improved.

The system control unit 11 is configured to accept and reply to the instructions transmitted during the test to complete the execution of various instructions, including but not limited to changing different power-off modes, test times, test cycles, and test result operations. The workload simulation unit 12 is coupled to the system control unit 11 to execute various workloads to be tested on the storage device 20 in accordance with instructions. The workload simulation unit 12 is mainly used to describe the ratio of behaviors and instructions that the NVMe storage device 20 can occur at the user end (according to the test specification JDEC 218, 219), such as continuous reading and writing, random reading and writing, reading and writing large files or small Files, etc. The content of the workload depends on the location of the storage device 20, such as word processing, game processing, or the workload of the mail server. The point is to simulate the behavior that occurs in the user's environment.

The power control unit 13 is connected to the system control unit 11 and can be selectively connected to the storage device 20 to be tested for performing power-off and power-on of the storage device 20 according to the command.

The protocol control unit 14 is also coupled to the system control unit 11 and is selectively connectable to the storage device 20 to be tested. The protocol control unit 14 can be used to send relevant test protocols, such as command/process power-off test, initial power-off test, pressure test related protocol specifications and procedures, for performing the inspection, comparison and judgment test of the storage device 20 test. Whether or not.

The data temporary storage unit 15 is connected to the system control unit 11 to record key data and information, so that the protocol control unit 14 checks the correctness of the data stored in the storage device 20, such as the length and position of the N data written before the power failure and other A fixed data area that will not be covered, etc.

Thereby, the group constitutes an abnormal power-off test device for a storage device that can be programmed and simulates an abnormal power-off test.

The test equipment 10 is in accordance with the test specifications of the NVMe storage device 20 (e.g., JDEC 218, 219) in actual use. The system control unit 11 of the test equipment 10 operates the storage device 20 according to the parameters of the workload simulation unit 12, and uses the power control unit 13 to perform an abnormal power-off action. The workload simulation unit 12 contains the ratio of all instructions that simulate various workload environments to the occurrence of each instruction. The protocol control unit 14 then performs various test modes on the storage device 20, including but not limited to the command/flow power down test, the initial power down test, and the stress test.

The second and third figures show a flow chart of the method for testing the abnormal power failure of the storage device of the present invention.

First, step S101 is executed to enable at least one storage device to perform operation of at least one simulated workload. The system control unit 11 of the test device 10 causes the storage device 20 to perform at least one simulated workload, such as continuous reading and writing, random reading and writing, reading and writing large files or small files, according to the setting of the workload simulation unit 12.

Then, in step S102, the storage device 20 performs a power-off test including a specified test mode according to the setting. The power control unit 13 performs the power-off of at least one specified time during the simulated workload operation of the storage device 20 according to the setting. The protocol control unit 14 performs a one-test mode corresponding to the power-off of the storage device 20. The test mode can be an instruction/process power down test A, a device initialization power down test B, or a stress test C. The instruction/process power-off test A includes an instruction power-off test or a process power-off test consisting of at least one command power-off test. The test mode may further be a stress test C that repeatedly executes the instruction/flow power failure test A or the device initialization power failure test B. If the trigger device initializes the power-off test B after the command/flow power-off test A is powered off, it is determined whether the power is continuously disconnected or powered according to the pressure test condition of the device initialization power-off test B. If the device initialization power-off test B is not entered, the normal upper current of step S103 is continued. Cheng. The execution flow for different test modes is shown in the third figure.

When the command/flow power failure test A is executed, after step S101, step S1022 is executed, and the command to power off in the middle of the power is turned off according to the setting. At the stage after the storage device 20 is powered up to when the protocol control unit 14 can command, the workload simulation unit 12 causes the storage device 20 to execute the simulated workload for a period of time. Then, the power control unit 13 is powered off in accordance with the ratio of the command. In order to increase or decrease the frequency at which a particular instruction occurs, the programmable parameters of the protocol control unit 14 are added herein to replicate the behavior of a particular situation or any analog NVMe instruction and to power down the instructions of the storage device 20. The instructions may include instructions specified by the NVMe protocol. In the process of power failure, the issued NVMe instruction does not have to wait for the response of the instruction, but performs power-off at any time after the instruction is sent. Such a sudden power-off can effectively attack the device firmware being moved or written. The action of the flash memory 22.

Next, step S1023 is executed to determine whether to perform the device initialization power-off test. After the power is turned off, it is judged according to the test parameters whether or not the device initialization power-off test B is performed. If the parameter is set to perform the device initializing power-off test B, step S1026 is performed. On the other hand, if the parameter is set to not require the device initialization power-off test B, step S103 is performed.

When the execution device initializes the power-off test B, after step S1023, step S1026 is executed, and the storage device is powered on. This process must be performed after a power outage.

Then, step S1027 is executed to set CC.EN to 1. According to the NVMe protocol specification, when the test device 10 temporarily sets the CC.EN of the NVMe storage device 20 to 1, the test device 10 repeatedly reads the CSTS.RDY state in the temporary storage data of the storage device 20 until it is converted from 0 to 1. until. In the converted project, the storage device 20 must comply with the timeout value of the device control unit 21. When converted to 1, the storage device 20 is ready to be completed, the test device 10 can further The relevant instructions are stored on the storage device 20.

Then, in step S1028, the test system is powered off during the average initialization time. During the CSTS.RDY conversion process, the device firmware may be processing the previously unfinished tasks (possibly GC, WL, or POR, etc.), or accessing or moving the data of the flash memory 22. Adding programmable parameters at this time and performing a power outage in the process will have a higher probability of hitting the device firmware to handle the complexities of various complex tasks.

Next, step S1029 is performed to determine whether the device initialization power-off test condition is satisfied. The storage device 20 determines whether the conditions of the specified test are completed according to the test specification. If the specified test condition is not satisfied, step S1026 is re-executed, and the test of the setting condition of the storage device 20 is continued. If the specified test condition has been met, step S103 is performed.

In the stress test C, the command/flow power-off test A and the device initial power-off test B are integrated into the overall power-off system, and the stylized conditions are added as the basis for the pressure test stop. In this way, various possible factors can be combined into one system to achieve the purpose of continuous repeated triggering.

Then, step S103 is executed to restart the storage device 20 and check whether the initialization time of the storage device 20 is normal. After the storage device 20 is powered off at a specified time during operation, the power control unit 13 re-powers the storage device 20. The protocol control unit 14 checks whether the initialization time of the storage device 20 is normal according to the test specification. If it is normal, step S104 is performed. If it is not normal, it means the test failed.

Next, step S104 is performed to determine whether the pressure test condition of the specified command power-off test is satisfied. After the storage device 20 completes the test of the command/flow power-off test A or the device initialization power-off test B, it is determined according to the test specification whether the corresponding command power-off test condition in the specified test is completed. If the test is not met In the test condition, step S102 is executed again, and the storage device 20 is tested for the setting conditions. If the specified command power-off test condition has been met, step S105 is performed.

Thereafter, step S105 is performed to compare whether the key data of the storage device is correct. After determining that the initialization time of the re-powering operation of the storage device 20 is normal, the system control unit 11 further compares the correctness of the key data stored in the storage device 20 through the data temporary storage unit 15, for example, the N pen written before the power failure. The length and location of the data and other fixed data areas that will not be covered. If the key data is correct, step S106 is performed. If the critical data is incorrect, the test failed.

Finally, step S106 is performed to check whether the storage device 20 satisfies the test condition. After the storage device 20 sequentially completes the test of the specified power-off test mode, the check of the initialization time, and the comparison of the key data of the stored data, it further determines whether the specified test mode, the number of tests, and the test cycle condition are completed according to the test specification. If the specified test condition has not been met, step S101 is performed again, and the storage device 20 is tested for the set condition. If the specified test conditions have been met, the test is successful and the test of the storage device 20 is terminated.

According to the principle description of the overall power-off test of the present invention, after jumping to the next stage after power-off, it is necessary to check whether the initialization time of the CSTS.RDY of the NVMe storage device 20 after normal power-on is in the timeout of the NVMe protocol specification (timeout) ) completed in time. If the inspection fails, the device firmware may have a problem under the influence of the previous power failure, and the damage caused by the power failure behavior to the device firmware must be re-examined and repaired.

Through the above description, the present invention has several advantages and practical values for testing the NVMe storage device 20 through the design of the abnormal power-off test device and its abnormal power-off test method.

First, the present invention can be used to perform a specified test by using an instruction power-off test. test. Different from the conventional random power-off test method, the present invention can control the power-off command so that it can meet the test requirements at a certain time after the command is sent.

Furthermore, the process power-off test of the present invention can perform a programmable test design, allowing the test to control various NVMe behaviors according to the NVMe protocol, and programmatically control the power-off to occur in various NVMe behaviors. Control the occurrence of a specific behavioral outage.

The stress test of the present invention can be applied to the command/flow power-off test and the initial power-off test to increase the random situation simulation or test cycle times of the workload, thereby increasing the number of error types to verify the device firmware in the storage device 20. The proper rate of handling different types of errors.

In addition, the present invention is tested by a programmable test design, and the simulation is performed under abnormal power failure to quickly discover potential problems of the device firmware and algorithmic loopholes, and improve the stability of the device firmware of the storage device. Reliability.

In summary, it can be understood that the present invention has excellent creativity, effectively solves existing problems, greatly enhances efficacy, and does not see the same or similar product creation or public use in the same technical field. Therefore, the present invention meets the requirements for "novelty" and "progressiveness" of the invention patent, and is an application for filing an invention patent according to law.

Claims (7)

A test device, according to at least one test specification, simulating at least one workload in at least one storage device, and performing at least one test at a specified time, the test device comprising: a system control unit capable of transmitting, receiving, and responding to a test a work load simulation unit connected to the system control unit and capable of performing various workload simulations to be tested on the storage device according to the instruction; a power control unit connected to the system control unit and selectively connecting the storage a device, and performing power-off and power-on of the storage device according to an instruction; a protocol control unit, connecting the system control unit, and selectively connecting the storage device, and transmitting an instruction of a relevant test protocol; a data temporary storage unit, The system control unit is connected, and key data and information can be recorded to check the correctness of the stored data of the storage device, wherein the test device is programmable. The test device of claim 1, wherein the storage device is a storage device of NVMe. A test method for simulating at least one workload in at least one storage device according to at least one test specification, and performing at least one test at a specified time, the test method comprising the steps of: (S101), causing at least one storage device to perform at least one Simulating the operation of the workload, and then performing the step (S102); In the step (S102), the storage device performs at least one power-off test according to the setting, the power-off test mode includes at least one command/flow power-off test, one device initial power-off test or a stress test, and then the step is performed (S103) The step (S103), re-powering the storage device, and checking whether the initialization time of the storage device is normal, if it is normal, executing the step (S104), if not, the test fails; the step (S104), whether Satisfying the stress test condition of the specified command power-off test, when the test condition has not been met, the step is re-executed (S102), and if the specified command power-off test condition has been met, the step (S105) is continued; the step (S105) Whether the key data of the storage device is correct, if yes, the step (S106) is performed, if not, the test fails; and the step (S106), checking whether the storage device meets the test condition, and if the test condition is not met, Then return to the execution of this step (S101) to continue the test, if the conditional test condition is met, the test is successful and the test is ended. The test method of claim 3, wherein the command/process power-off test can be an instruction power-off test or a process power-off test consisting of at least one command power-off test. The test method of claim 3 or 4, wherein the test mode may be a stress test of repeatedly executing the instruction/flow power failure test or the device initial power failure test. The test method of claim 3, wherein in the step (S102), after the power is cut off, the organic rate triggers the initial power-off test module, and according to Initialize the power failure test condition to determine whether to continuously power off. If the initial power failure test is not entered, continue the normal power-on process of this step (S103). The test method of claim 3, wherein the storage device is a storage device of NVMe.