A NAND flash controller multi-command hybrid test method

Abstract

The application discloses a NAND Flash controller multi-command mixed test method, which can arbitrarily adjust the weight of various types of commands, and the probability of each command appearing in actual operation is basically consistent with the preset weight value, thereby solving the problem that the traditional test method only tests a single command type each time, the parameter modification is not flexible enough, and the interaction between various commands in a complex real firmware application environment cannot be simulated. A dynamic management mechanism of the memory blocks in the test range is realized through TLC / SLC available block grouping, command suspension list and full-written block list, so that long-time test of the automation script is facilitated, and the maintenance cost of the tester is reduced. Data verification is performed in real time, and sample data with overall randomness and special detection points inserted in key positions are adopted, so that the accuracy and operation efficiency are considered.

Description

Technical Field

[0001] This invention relates to the field of NAND Flash testing, specifically a method for mixed testing of multiple commands for NAND Flash controllers. Background Technology

[0002] NAND flash memory is a common non-volatile storage medium with advantages such as fast access speed, high storage density, and low cost per unit capacity, making it widely used in solid-state storage devices (SSDs). However, its erase-before-write, block-by-block erase, word-by-word (WL) write, and page-by-page read characteristics mean it cannot perform random addressing like Nor flash memory. It requires more complex control logic to satisfy the I / O access protocol, necessitating NAND flash controller hardware and corresponding drivers. A NAND flash controller has multiple channels connected to multiple flash memory chips, operating independently and in parallel. It can execute various NAND flash operation commands, including erase, write (program), read, and set / get feature commands. The specific configuration information of these commands is stored in the SRAM of each channel of the controller, called the command slot.

[0003] Existing NAND Flash control testing methods are limited in function and have low command type coverage, lacking effective means to verify the reliability of NAND Flash controller hardware and its drivers in complex environments with multiple commands mixed in batches.

[0004] Existing NAND Flash control testing methods generally test a single command type. For testing multiple commands, there are two approaches: one is through a simulation platform, which can only handle scenarios with dozens of commands mixed together and cannot adapt to complex environments with multiple commands mixed in batches; the other is a simple mixture of multiple test commands, which simply superimposes the test methods of various command types. The parameter modification is not flexible enough and cannot simulate the interaction between multiple commands in complex and real firmware application environments. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-command hybrid testing method for NAND Flash controllers. Under the condition of meeting execution efficiency requirements, it supports random parallel testing of multiple NAND Flash commands, thereby improving test coverage and verification results.

[0006] To solve the aforementioned technical problem, the present invention adopts the following technical solution: a multi-command mixed testing method for NAND Flash controllers, comprising the following steps: S01) The NFC test program obtains the test case configuration parameters from the command receiving and parsing module and checks the validity of the parameters; S02) Determine the starting value and range of the test physical address based on the parameters. Starting from the starting physical block address, check whether there are enough consecutive available blocks within the test range. If the number is sufficient, proceed to step S03). If the number is insufficient, return an error message indicating insufficient resources. S03) The NFC test program groups the available blocks selected in step S02) and uses a portion of them as TLC storage blocks and a portion as SLC memory according to the preset ratio in the test case configuration parameters. The erase command is used to clear the existing data on each available block in turn. S04) Randomly select some storage blocks within each group, generate sample data associated with the flash physical address and write it as the initial data source for read commands during the mixing process, establish a full list to record the full blocks, and select blocks from the list for subsequent read and erase commands. S05) Generate random numbers based on the weight ratio of each type of command in the test case configuration parameters, and determine the type of command to be generated based on the random numbers; repeat the random process to complete the remaining parameters of the command, and enter different command generation processing flows according to different command types; The remaining parameters of the command here include general parameters and parameters specific to each command. General parameters include command index number, physical address, memory data address, etc. In addition, different command types have their own specific parameters. For example, read commands have read modes such as 4K / 16K / 512, and write commands have options such as whether to include meta data. Completing the remaining parameters of the command is a common technique in this field, so it is not described in detail.

[0007] S06) Determine whether the generated command meets the sending conditions. If it does, send the command to the NFC driver. If it does not meet the conditions, temporarily add the command to the pending list. S07) Every certain period of time or when the number of sent commands or temporarily suspended commands reaches a set threshold, the message recycling mechanism is triggered. The status message returned by the NFC driver is received, data verification and error handling are performed, and the full list and the suspended list are updated. S08) The NFC test program provides statistical information on the types and quantities of completed commands at regular intervals. When the cumulative number of completed test commands reaches a preset value, or the test process is terminated due to an abnormal reason, the mixed test ends.

[0008] Furthermore, in step S04), the sample data associated with the flash physical address includes random padding data and specific value padding data. Specific value padding involves replacing the flash physical address with the offset number of the 512 / 520 data block in 4K at certain random positions in the header and middle of each 4K 512 / 520 data block, followed by indicating the next location to store flash physical address information. Random padding involves uniformly filling positions other than those specified by the specific value padding with random numbers.

[0009] Furthermore, random numbers are generated based on the weight ratios of each command type in the test case configuration parameters. The process of determining the command type to be generated based on the random numbers is as follows: the weight ratios of each command type are summed, and the sum is denoted by S. The NFC test program generates a random number in real time that is in the range [0, S-1] and follows a uniform distribution. The type of command to be generated and sent next is determined based on the specific value of the random number. If the random number is in the range [(W1 + ... + W...]...]... i-1 , W1 +…+ W i-1 +W i Between -1], select the type i command, W i This represents the proportional weight of command type i, where i ranges from [1, n] and n is the total number of command types.

[0010] Further, in step S06), for the erase command, firstly, a storage block that is already filled with data and whose type matches is randomly selected within the physical address range of the test coverage according to the TLC / SLC attribute. If the number of filled storage blocks is less than the set threshold, the erase command is skipped. If there is an unprocessed read command on the selected storage block, the erase command is temporarily added to the pending list and subsequent new read commands are prevented from being sent to the storage block. The block is released after the existing read commands on it are completed; otherwise, it is sent directly to the NFC driver.

[0011] In step S06), for a write command, a non-full storage block is randomly selected from the corresponding group according to the TLC / SLC classification. The Flash physical address is the next WL after the last write position of the block, and the memory data address points to the sample data prepared for the WL. If all storage blocks in the group are full, one or more storage blocks are randomly selected for erasure. The Flash physical address of the pen write command is set to the first WL of the selected storage block, and it is placed in the pending list. It is sent to the NFC driver after the erasure is completed.

[0012] Further, in step S06), for a read command, if it is a fake read without actual data transmission, then any page of any storage block within the test range is randomly assigned as its physical address; if it is a normal read command with actual data transmission, firstly, according to its TLC / SLC distinction, a page of the full storage block in the corresponding group is randomly selected as the physical address, and then the other attributes of the read command are completed by random lottery. Then, according to the above configuration, memory space is allocated for the command to store the data read from NandFlash and the command context including the physical address is recorded for subsequent data verification. After the memory allocation is successful and the context recording is completed, the command is sent to the NFC driver.

[0013] Further, in step S06), for Set / Get Feature type commands, the NFC test program maintains a list FeatureList that records the initial and current values ​​of each LUN Feature. At the start of the test, the initial value of each LUN Feature is obtained through the Get Feature command and used as the current value. During the mixed test, the Set / Get Feature command randomly selects any LUN within the test range as the physical address. If it is a Set Feature command, a Feature value within the legal value range is randomly selected as the parameter and the current value in FeatureList is updated. If it is a Get Feature command, it is compared with the current value in FeatureList after the command returns to confirm that the Set Feature command is valid. After the mixed test process is completed, the initial value recorded in FeatureList is used as the parameter to send the Set Feature command to all LUNs within the test range to restore the state before the test.

[0014] Further, after step S07) triggers the message recycling mechanism, the status message returned by the NFC driver is processed, and the corresponding processing flow is executed according to the message type: after the erase command is completed, the storage block where the command is located is removed from the full list; after the write command is completed, it is checked whether the storage block where the command is located is full; if it is full, it is added to the full list to allow subsequent access; after the read command is completed, the data is verified according to the context of the record, and then the memory resources requested by the command are released; in addition, it is checked whether the storage block where the read command is located is a command to be erased; if so, and there are no other unfinished read commands on the block, it is erased; the suspended list is traversed and commands that meet the conditions are executed; when the NFC driver returns an error status, the physical address where the error command is located is located and output, and the command is resent or the test is terminated according to the type and severity of the error.

[0015] Furthermore, the obtained test case configuration parameters include the test scope, the starting flash physical address used by the test program, the memory addresses used for writing sample data / receiving data read from Nand Flash, the proportion of each type of command, and the total number of mixed test commands.

[0016] The beneficial effects of this invention are: 1. The weights of various commands such as erase, write, read, and feature can be arbitrarily adjusted. The probability of each command appearing during actual runtime is basically consistent with the preset weight values. Command attribute parameters are highly randomized, which can serve as a behavioral model reference for FTL firmware in certain application scenarios, or be used to accelerate the reproduction of hardware bugs that can only be triggered under extreme conditions by the Nand Flash controller. Traditional testing methods only test a single command type each time, and parameter modification is not flexible enough, failing to simulate the interaction between multiple commands in complex and realistic firmware application environments.

[0017] 2. TLC / SLC can use block grouping, command pending list (PendingList), and filled block list (FilledList) to realize a dynamic management mechanism for storage blocks within the test range. The total number of commands can be specified in the preset parameters. After the hybrid test program starts running, there is no need for external intervention to detect the empty or full status of the storage blocks in the tested area. Moreover, space resources are recycled, which facilitates long-term testing by automated scripts and reduces the maintenance costs for testers.

[0018] 3. Data verification is performed in real-time. Each time a command reads data from the Nand Flash is completed, data verification is immediately performed based on the command context. This allows for the immediate interception of data anomalies, preventing errors from escalating and propagating. Furthermore, by using sample data with overall randomization and insertion of special test points at key locations, complete verification of the entire page's data is unnecessary, balancing accuracy and operational efficiency. Traditional methods execute batch data comparison commands uniformly after the test program ends, which is time-consuming and has a lag. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the Nand Flash testing system upon which this method is based; Figure 2 This is a flowchart of the method. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1 This embodiment discloses a multi-command mixed testing method for NAND Flash controllers, which is based on Figure 1The test system shown includes a command receiving and parsing module, a Nand Flash controller test program (hereinafter referred to as the NFC test program), a Nand Flash controller driver (hereinafter referred to as the NFC driver), and a Nand Flash controller hardware (hereinafter referred to as NFC). The command receiving and parsing module is responsible for providing test case configuration parameters for the NFC test program. The NFC driver and NFC are the objects under test, which are equivalent to black boxes. The main logic described in this method runs on the NFC test program, and it interacts with the NFC driver through command and status message FIFOs.

[0022] like Figure 2 As shown, the main steps of the testing method are as follows: 1. The NFC test program in the firmware of the board containing the Nand Flash controller parses the instructions from the host computer serial port input or the automated test script to obtain various parameters required for the test case to run. These parameters mainly include the test scope (multi-channel parallel, single-channel multi-LUN parallel, single-LUN test, and other different modes), the starting flash physical address (PBA) used by the test program, the memory addresses used for the sample data to be written and the data read from the Nand Flash, as well as the proportion of each type of command such as read, write, and erase, and the total number of mixed test commands.

[0023] 2. The NFC test program checks the validity of various parameters (mixing ratio, read / write memory address, etc.) of the obtained test cases. If they exceed the allowed range, it will report an error and prompt the valid parameter values ​​and the constraint relationship between variables.

[0024] 3. Based on the NandFlash storage organization structure (number of Channels, CEs, LUNs, Planes) and the defined test range, starting from the starting physical block address (Block number), check if there are enough consecutive available blocks (non-bad blocks) within the test range. If the condition is met, continue with the subsequent test steps; if there are not enough available blocks, return an error message indicating insufficient resources.

[0025] 4. The NFC test program groups the available blocks selected in the above process (3), and according to the preset ratio in the configuration parameters, one part is used as TLC storage blocks and the other part is used as SLC storage blocks. The corresponding erase command is used to clear the existing data on each available block in turn.

[0026] 5. Generate test sample data (Pattern) to be written in the given memory address. The contents of each Page and GU are generally filled with random numbers without any regularity. However, at certain random positions in the header and middle of each 512 / 520 data block of each 4K, the flash physical address (PBA, accurate to GU) and the offset number of the 512 / 520 data block in 4K are replaced. Then, the location where the above PBA information is stored is indicated.

[0027] 6. Randomly select a portion of the sample data generated in step (5) above from each of the grouped TLC / SLC storage blocks as the data source for the read command at the start of the mixing process (to prevent ECC errors caused by reading blocks that are not fully written). Create a list (FilledList) to dynamically record the blocks that have been fully written.

[0028] 7. After all storage blocks requiring pre-filled data have been written, the mixed testing process begins. The weighted proportions of each command type (W1, W2, ..., Wn) are summed (set to a value of S). The test program generates a random number in real-time, ranging from 0 to S - 1 and following a uniform distribution. Based on the specific value of the random number, it determines which test command to generate and send next (e.g., selecting type 1 command between 0 and W1 - 1, type 2 command between W1 and (W1 + W2) - 1, type i command between (W1 + ... + Wi - 1) and (W1 + ... + Wi - 1 + Wi) - 1, and so on, ensuring that the probability of each command type during the mixed testing process basically conforms to the preset proportion). After determining the type, different command generation and processing flows are initiated according to different command types such as read, write, and erase.

[0029] In this embodiment, the input command weight ratio values ​​are integers, not percentages, such as 3:5:10. The calculation is to sum the weight ratios of each type of command, i.e., S=3+5+10, and then proceed with the subsequent processes.

[0030] 8. For erase commands, firstly, based on their TLC / SLC attributes, randomly select a memory block that is already filled with data and matches the data type within the physical address range of the test coverage. If the number of filled memory blocks is less than the set threshold, skip this erase command (since erasing is done by block and reading / writing is done by page, the number of read / write commands is generally much greater than the number of erase commands). If there is an unprocessed read command on the selected memory block, temporarily add the erase command to the PendingList and prevent subsequent new read commands from being sent to the memory block. Release the erase command after the existing read command on the block is completed; otherwise, send it directly to the NFC driver.

[0031] 9. For write (programming) commands, a non-full storage block is randomly selected from the corresponding group according to the TLC / SLC classification. The Flash physical address is the next WL after the last write position of the block (for Multiplane type operations, a constraint condition is added that the block address is a multiple of the number of flash memory particles). The memory data address points to the sample data prepared for the WL. If all storage blocks in the group are full, one (or multiple in Multiplane) storage block is randomly selected for erasure as described in step (8). The Flash physical address of the pen write command is set to the first WL of the selected storage block and placed in the PendingList. It is sent to the NFC driver after erasure is completed.

[0032] 10. For read commands, if it is a dummy read (used for refresh to prevent excessively high BER) without actual data transmission, then any page of any storage block within the test range is randomly assigned as its physical address (regardless of whether the block has been written to). If it is a normal read command with actual data transmission, firstly, based on its TLC / SLC distinction, a page of a full storage block (located in the FilledList) in the corresponding group is randomly selected as the physical address. Then, other attributes of the read command are completed by random selection (including Single / MultiPlane type, number of GUs read, sector size (4096 / 4104 / 512 / 520), data transmission method (direct memory address or DMA Link descriptor, etc.), whether it contains meta data, and which priority command slot of the NandFlash controller to send it to, etc.). Then, according to the above configuration, memory space is allocated for the command to store the data read from the NandFlash and the command context, including the physical address, is recorded for later data verification (if the currently available memory resources are insufficient, the read command is first added to the PendingList). After successfully allocating memory and completing context recording, the command is sent to the NFC driver.

[0033] 11. For Set / Get Feature commands (setting the Nand threshold voltage Vth during Read Retry), the NFC test program maintains a list (FeatureList) recording the initial and current values ​​of each LUN Feature. At the start of the test, the Get Feature command retrieves the initial Feature value of each LUN and uses it as the current value. During the mixed test, the Set / Get Feature command randomly selects any LUN within the test range as the physical address. If it is a Set Feature command, it randomly selects a Feature value within the valid value range as the parameter and updates the current value in the FeatureList, then sends it to the NFC driver. If it is a Get Feature command, it compares the current value with the current value in the FeatureList after the command returns to confirm the validity of the Set Feature command. After the mixed test process ends, the initial value recorded in the FeatureList is used as the parameter to send the Set Feature command to all LUNs within the test range to restore the state before the test.

[0034] Here, "Feature" refers to specific physical attribute values ​​within a Nand Flash chip, such as the threshold voltage Vth used when reading data from a flash cell. These values ​​can be adjusted and viewed using external commands. "Set Feature" modifies these attributes, while "Get Feature" retrieves them.

[0035] 12. During the mixed testing process, at regular intervals, or when the number of sent commands and temporarily suspended commands (located in the PendingList) reaches a certain threshold, a message recycling mechanism is triggered to process the status messages returned by the NFC driver. The corresponding processing flow is executed based on the message type: after an erase command is completed, the storage block containing the command is removed from the FilledList; after a write (programming) command is completed, it is checked whether the storage block containing the write command is full; if full, it is added to the FilledList to allow subsequent read access; after a read command is completed, data verification is performed based on the recorded context, and then the memory resources requested by the command are released. Additionally, it is checked whether the storage block containing the read command is in a state awaiting erasure; if so, and there are no other incomplete read commands on that block, it is erased. The PendingList is traversed, and commands that meet the conditions are executed. When the NFC driver returns an error status, the physical address of the erroneous command is located and output, and the type and severity of the error determine whether to resend the command or terminate the test.

[0036] 13. The NFC test program provides statistical information on the types and quantities of completed commands at regular intervals to allow testers to understand the progress. When the cumulative number of completed test commands reaches the preset value, or when the test process is terminated due to abnormal reasons (including command reception timeout or NFC driver returning an error that cannot be processed), the mixed test ends, the resources requested during the process are released, and the final test report is output.

[0037] The above description is merely the basic principle and preferred embodiment of the present invention. Improvements and substitutions made by those skilled in the art based on the present invention are within the scope of protection of the present invention.

Claims

1. A method for testing multiple commands mixed in a NAND Flash controller, characterized in that: Includes the following steps: S01) The NFC test program obtains the test case configuration parameters from the command receiving and parsing module and checks the validity of the parameters; S02) Determine the starting value and range of the test physical address based on the parameters. Starting from the starting physical block address, check whether there are enough consecutive available blocks within the test range. If there are, proceed to step S03). If not, return an error message indicating insufficient resources. S03) The NFC test program groups the continuous available blocks selected in step S02) into groups, and according to the preset ratio in the test case configuration parameters, a part is used as a TLC storage block and a part is used as an SLC memory. The erase command is used to clear the existing data on each continuous available block in turn. S04) Randomly select some storage blocks within each group, generate sample data associated with the flash physical address and write it as the initial data source for read commands during the mixing process, establish a full list to record the full storage blocks, and select the full storage blocks in the list as the physical address for subsequent read and erase commands. S05) Generate random numbers based on the weight ratio of each type of command in the test case configuration parameters, and determine the type of command to be generated based on the random numbers; The repeated random process completes the remaining parameters of the command, and enters different command generation and processing flows according to different command types; S06) Determine whether the generated command meets the sending conditions. If it does, send the command to the NFC driver. If it does not meet the conditions, temporarily add the generated command to the pending list. S07) Every certain period of time or when the number of sent commands and temporarily suspended commands reaches a set threshold, the message recycling mechanism is triggered, the status message returned by the NFC driver is received, data verification and error handling are performed, and the full list and the suspended list are updated. S08) The NFC test program provides statistical information on the types and quantities of completed commands at regular intervals. When the cumulative number of completed test commands reaches a preset value, or the test process is terminated due to an abnormal reason, the mixed test ends.

2. The NAND Flash controller multi-command mixed testing method according to claim 1, characterized in that: In step S04), the sample data associated with the flash physical address includes random padding data and specific value padding data. Specific value padding is to replace the flash physical address with the offset number of the 512 / 520 data block in 4K at certain random positions in the header and middle of each 512 / 520 data block of each 4K storage block, and then indicate the position where the next flash physical address information is stored. Random padding is to fill the positions outside the specific value padding with random numbers evenly.

3. The NAND Flash controller multi-command mixed testing method according to claim 1, characterized in that: Random numbers are generated based on the weight ratios of each command type in the test case configuration parameters. The process of determining the command type to be generated based on the random numbers is as follows: the weight ratios of each command type are summed, and the summation value is represented by S. The NFC test program generates a random number in real time that is in the range [0, S-1] and follows a uniform distribution. The type of command to be generated and sent is determined based on the specific value of the random number. If the random number is between (W1+…+ Wi-1) ~ (W1+…+ Wi-1 + Wi) - 1, command type i is selected, where Wi represents the weight ratio of command type i, and the value of i is in the range [1, n], where n is the total number of command types.

4. The NAND Flash controller multi-command mixed testing method according to claim 1, characterized in that: In step S06), for an erase command, firstly, a storage block that is already filled with data and whose type matches the TLC / SLC attribute is randomly selected within the physical address range of the test coverage. If the number of filled storage blocks is less than the set threshold, the erase command is skipped. If there is an unprocessed read command on the selected storage block, the erase command is temporarily added to the pending list, and subsequent new read commands are prevented from being sent to the storage block. The block is released after the existing read commands on it are completed; otherwise, it is sent directly to the NFC driver.

5. The NAND Flash controller multi-command mixed testing method according to claim 1, characterized in that: In step S06), for a write command, a non-full storage block is randomly selected from the corresponding group according to the TLC / SLC classification. The Flash physical address is the next WL after the last write position of the non-full storage block, and the memory data address points to the sample data prepared for the WL. If all storage blocks in the group are full, one or more storage blocks are randomly selected for erasure. The Flash physical address of the pen write command is set to the first WL of the selected storage block, and it is placed in the pending list. It is sent to the NFC driver after the erasure is completed.

6. The NAND Flash controller multi-command mixed testing method according to claim 1, characterized in that: In step S06), for a read command, if it is a fake read without actual data transmission, then any page of any storage block within the test range is randomly assigned as its physical address; if it is a normal read command with actual data transmission, firstly, according to its TLC / SLC distinction, a page of the full storage block in the corresponding group is randomly selected as the physical address, and then the other attributes of the read command are completed by random lottery. Then, according to the above configuration, memory space is allocated for the read command to store the data read from NandFlash and the command context including the physical address is recorded for subsequent data verification. After the memory allocation is successful and the context recording is completed, the read command is sent to the NFC driver.

7. The NAND Flash controller multi-command mixed test method according to claim 1, characterized in that: In step S06), for Set / Get Feature type commands, the NFC test program maintains a list FeatureList that records the initial and current values ​​of each LUN Feature. At the start of the test, the initial value of each LUN Feature is obtained through the Get Feature command and used as the current value. During the mixed test, the Set / Get Feature command randomly selects any LUN within the test range as the physical address. If it is a Set Feature command, a Feature value within the legal value range is randomly selected as the parameter and the current value in FeatureList is updated. If it is a Get Feature command, the current value in FeatureList is compared with the current value after the command returns to confirm that the Set Feature command is valid. After the mixed test process is completed, the initial value recorded in FeatureList is used as the parameter to send the Set Feature command to all LUNs within the test range to restore the state before the test.

8. The NAND Flash controller multi-command mixed test method according to claim 1, characterized in that: Step S07) After triggering the message recycling mechanism, process the status message returned by the NFC driver and execute the corresponding processing flow according to the message type: After the erase command is completed, remove the storage block where the command is located from the full list. After the write command is completed, check whether the storage block where the write command is located is full. If it is full, add it to the full list to allow subsequent access. After the read command is completed, perform data verification according to the record context, and then release the memory resources requested by the command. In addition, check whether the storage block where the pen read command is located is a command to be erased. If so, and there are no other unfinished read commands on the storage block, erase it. Traverse the suspended list and execute the commands that meet the conditions. When the NFC driver returns an error status, locate and output the physical address where the error command is located, and decide whether to resend the command or terminate the test according to the type and severity of the error.

9. The NAND Flash controller multi-command mixed testing method according to claim 1, characterized in that: The obtained test case configuration parameters include the test scope, the starting flash physical address used by the test program, the memory addresses used for writing sample data / receiving data read from Nand Flash, the proportion of each type of command, and the total number of mixed test commands.

Images