JP2012238259A - Control unit, storage device and control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance data reliability of non-volatile memory and increase speed and stability of reading access time.SOLUTION: Data stored in non-volatile memory are estimated whether refreshing is required in near future; i.e. whether refreshing timing is coming close. This estimation uses a piece of estimation information which stores the number of bit errors at every access. When a piece of data is determined that the refreshing timing of which is close, a piece of clone data identical to original data is generated in the non-volatile memory using the data as the original data. In this case, a piece of management information is updated so that either one of the original data and the clone data should be a target of the access.

Description

  The present disclosure relates to a control device, a storage device, and a read control method, and more particularly to a technique for improving the reliability of a non-volatile memory and speeding up and stabilizing an access time.

JP 2009-70098 A JP 2007-334852 A JP 2007-193838 A JP 2007-58840 A

For example, a storage device using a non-volatile memory such as a NAND flash memory has become widespread. Nonvolatile memories are used in, for example, memory cards used in various electronic devices and information processing apparatuses, SSDs (Solid State Drives), eMMCs (Embedded MultiMedia Cards), and the like.
Patent Documents 1 to 4 disclose a storage device using a flash memory.

In a nonvolatile memory, a physical address is used as an address of a physical storage area. Thereby, a physical block, a physical page, and a physical sector are set. A physical page is composed of a plurality of physical sectors, and a physical block is composed of a plurality of physical pages.
Erasing (erasing) is performed in units of physical blocks, and writing (programming) and reading (reading) are possible in units of physical pages.
A logical address is used for address designation from the host side or the memory control unit side. Logical blocks, logical pages, and logical sectors based on logical addresses are associated with the physical addresses. As a result, when an access request is made, the logical address is converted into a physical address, and access to the actual flash memory is executed.

By the way, in the NAND flash memory, the reliability of data has been improved by using an error correction code in preparation for the occurrence of bit corruption (bit error) when the written data is read.
However, in recent years, due to the process miniaturization of NAND flash memory, the phenomenon that data is garbled easily occurs when the written data is repeatedly read repeatedly, and garbled bits that cannot be corrected occur. It's getting easier.

In such a situation, there is known a method for making a situation in which bit corruption is unlikely to occur by performing a data refresh operation while not only correcting a bit corruption but also correcting it.
In the conventional refresh method, when the number of occurrences of bit corruption, the number of times of reading or rewriting exceeds a certain threshold, it is determined that a refresh opportunity has been reached, and a refresh operation is performed.

  However, when such a conventional method is used, a refresh operation is started immediately when a state exceeding the threshold value is generated, so that the processing speed at the time of read access is lowered. Also, the processing speed varies depending on whether or not the refresh operation occurs.

  In view of such a problem, the present disclosure aims to increase the reliability of data stored in a nonvolatile memory and to increase the processing speed and stabilize the access speed.

  The control device of the present disclosure is data stored in a non-volatile memory, and for data determined to be close to a refresh opportunity based on information for evaluation regarding refresh necessity, the data is used as original data. A clone generation processing unit that performs processing for generating the same clone data as the original data in the nonvolatile memory, and so that either the original data or the clone data becomes an access target at the time of access execution. A management information processing unit for updating the management information; and an evaluation information processing unit for generating the evaluation information for the data when accessing the data stored in the nonvolatile memory.

  The storage device according to the present disclosure includes a nonvolatile memory and data stored in the nonvolatile memory, the data determined to be close to each other based on evaluation information regarding the necessity for refresh. Using the data as original data, the clone generation processing unit that performs processing for generating the same clone data as the original data in the nonvolatile memory, and either the original data or the clone data is A management information processing unit that updates management information so as to be an access target, and an evaluation information processing unit that generates the evaluation information for the data when the data stored in the nonvolatile memory is accessed .

  The control method of the present disclosure is data stored in a non-volatile memory, and for data for which it is determined that a refresh opportunity is close based on information for evaluation regarding refresh necessity, the data is used as original data. A clone generation step for generating the same clone data as the original data in the non-volatile memory, and so that either the original data or the clone data is an access target at the time of access execution, A management information processing step for updating the management information; and an evaluation information processing step for generating the evaluation information for the data when the data stored in the nonvolatile memory is accessed.

In the techniques of the present disclosure, the reliability is not a case where a refresh operation is required (such as when a refresh opportunity is reached) such as when a value representing the reliability of stored data exceeds a certain threshold value. Data for which a refresh operation is likely to be necessary is estimated by using the evaluation information storing the value indicating the property. Then, by replicating the clone data for the data in advance, a decrease in reliability is prevented without reducing the processing speed at the time of reading.
That is, clone data is duplicated for data that is expected to be required soon after a refresh operation, and data having the same content is held twice. One of the original data and the clone data is managed as main data to be accessed, and the other is set as sub data that is not to be accessed. When reading the data, the main data (for example, clone data) is read, and if the reliability is reduced due to repeated reading, the main data is logically managed without refreshing at that time. And sub data are exchanged, and new main data (for example, original data) is read out.
This eliminates the need for an immediate refresh operation even when reliability is reduced during reading.

According to the technique of the present disclosure, clone data is generated for data whose reliability tends to be relatively low, and the data contents are held twice, so that the reliability of the data can be improved. In particular, if the clone data side is the target of read access as the main data, the new data is the same as the refresh, and the reliability is restored.
Furthermore, when the refresh opportunity is reached for the main data that has been read, the main data and the sub data are switched, and thereafter, the new main data is read. As a result, the opportunity to perform the refresh operation can be made very rare, and in most cases, the access speed can be increased and stabilized without refresh.

It is a block diagram of the composition of the memory card of an embodiment of this indication. It is explanatory drawing of a function structure of the control part of embodiment. It is explanatory drawing of the case where refresh is performed by the conventional method, and a reliability fall. It is explanatory drawing of the fluctuation | variation of the access time by execution of refresh. It is explanatory drawing of stabilization of the access time in embodiment. It is explanatory drawing of the operation | movement accompanying the clone data production | generation of 1st, 2nd embodiment. It is explanatory drawing of an example of the information for evaluation of embodiment. It is a flowchart of the process at the time of starting of 1st Embodiment. It is a flowchart of the process at the time of the read of 1st Embodiment. It is explanatory drawing of the refresh opportunity proximity | contact estimation of embodiment. It is explanatory drawing of the data read count by operation | movement of embodiment. It is explanatory drawing of the update of the information for evaluation of 2nd Embodiment. It is a flowchart of the process at the time of starting of 2nd Embodiment. It is a flowchart of the process at the time of the read of 2nd Embodiment. It is explanatory drawing of operation | movement accompanying the clone data production | generation of 3rd Embodiment.

Hereinafter, embodiments will be described in the following order. Note that the memory card 1 shown in the embodiment is an embodiment of a storage device referred to in the claims. Further, the control unit 11 in the memory card 1 is an embodiment of the control device referred to in the claims, and the processing by the control unit 11 is an embodiment of the control method referred to in the claims.

<1. Memory card configuration>
<2. Reduction in processing speed and variation due to refresh>
<3. First Embodiment>
[3-1: Operation overview]
[3-2: Information for evaluation]
[3-3: Processing example]
<4. Second Embodiment>
<5. Third Embodiment>
<6. Modification>

<1. Memory card configuration>

FIG. 1 shows a configuration example of the memory card 1 according to the embodiment.
The memory card 1 is connected to the host device 2 and used as a storage device. Examples of the host device 2 include various electronic devices and information processing apparatuses such as personal computers, digital still cameras, video cameras, audio players, video players, game devices, mobile phones, and information terminals such as PDAs (Personal Digital Assistants). is assumed.

  The memory card 1 includes a control unit (CPU: Central Processing Unit) 11, an SRAM (Static Random Access Memory) 12, a device interface 13, a buffer RAM, and a non-volatile memory (Non-Volatile Memory) 15.

The control unit 11 controls the entire memory card 1. Therefore, the control unit 11 sequentially executes instruction codes placed in the SRAM 12. The control unit 11 mainly executes data writing and reading in accordance with commands from the host device 2. Therefore, the control unit 11 performs data transmission / reception operation control of the device interface 13 with the host device 2, control of write / read operation of the buffer RAM 14, and control of access operation to the nonvolatile memory 15.
The SRAM 12 is used as a storage of a program (firmware) executed by the control unit 11 and a work area.
The device interface 13 performs communication with the host device 2.
The buffer RAM 14 is used for buffering transfer data (write data and read data) with the host device 2.
The nonvolatile memory 15 is, for example, a NAND flash memory.

As a basic operation of the memory card 1, at the time of data writing, a write request, a data size, and data to be written are sent from the host device 2 together with a write request.
Data to be written sent from the host device 2 is received by the device interface 13 and buffered in the buffer RAM 14. Then, data is written into the nonvolatile memory 15 under the control of the control unit 11. The control unit 11 controls these operations according to the write request, write address, and data size.
When reading data, a read address and data size are sent from the host device 2 together with a read request. The control unit 11 reads the data instructed from the nonvolatile memory 15 based on the read address and the data size, and buffers the data in the buffer RAM 14. The control unit 11 also performs error correction processing on the buffered read data. The read data is transferred from the buffer RAM 14 to the device interface 13 and transmitted to the host device 2.

By the way, the non-volatile memory 15 uses a physical address as an address of a physical storage area. Thereby, a physical block, a physical page, and a physical sector are set. A physical page is composed of a plurality of physical sectors, and a physical block is composed of a plurality of physical pages.
Erasing (erasing) is performed in units of physical blocks, and writing (programming) and reading (reading) are possible in units of physical pages.
A logical address is used for address designation from the host device 2 side. Logical blocks, logical pages, and logical sectors based on logical addresses are associated with the physical addresses. That is, the control unit 11 has management information for associating a logical address with a physical address, and when an access request is made from the host device 2, converts the designated logical address into a physical address with reference to the management information. Then, the actual nonvolatile memory 15 is accessed using the physical address.

Here, in the case of the present embodiment, as a function (a processing function realized by software) for executing an operation to be described later, as shown in FIG. 2, an access control unit 31, a clone generation processing unit 32, a management information processing Unit 33, evaluation information processing unit 34, and refresh control unit 35.
In addition, although it is assumed here that these functional parts are provided as software functions that are manifested as processing by the CPU 11, these may be formed by hardware.
Further, under the management of the control unit 11, for example, an area as a management information unit 41 and an evaluation information table 42 is prepared in the SRAM 12, and management information and evaluation information are stored.

The management information unit 41 stores management information such as the above-described logical address / physical address conversion information and directory information of data stored in the nonvolatile memory 15. That is, the information is necessary for the control unit 11 to execute writing to and reading from the nonvolatile memory 15 in accordance with a command from the host device 2.
In the management information, one of the original data and the clone data is managed as an access target for data in which clone data described later is generated. In other words, one physical address of the original data and the clone data is derived corresponding to a certain logical address. The other data is managed as data that is not an access target and is the same data as the access target data.

  As will be described in detail later, the evaluation information table 42 stores evaluation information as information for evaluating the degree of reliability for each data unit (physical block unit as an example). In the case of the present embodiment, the evaluation information is information for estimating whether or not each data unit will reach a refresh opportunity in the near future. For example, the information includes at least one of the number of bit error occurrences, the number of reads, and the number of erasures for each predetermined data unit.

One or both of the management information and the evaluation information may be always stored in the nonvolatile memory 15, or stored in the nonvolatile memory 15, expanded in the SRAM 12 at the time of startup, and referenced / updated by the control unit 11. You may be made to do. Alternatively, a nonvolatile storage area other than the nonvolatile memory 15 may be prepared and stored therein.
That is, the management information and the evaluation information may be held in any form as long as they are held in a nonvolatile manner and can be referred to / updated by the control unit 11 while the memory card 1 is activated.

  The access control unit 31 in the control unit 11 has a function of executing and controlling data write access or data read access to the nonvolatile memory 15 in accordance with a command from the host device 2. The access control unit 31 refers to the management information in the management information unit 41 and executes access to the nonvolatile memory 15 corresponding to the command from the host device 2.

The clone generation processing unit 32 stores the data stored in the nonvolatile memory 15 and determines that the refresh opportunity is close based on the evaluation information in the evaluation information table 42 regarding the need for refresh. This function is a function for generating the same clone data as the original data in the nonvolatile memory 15 using the data as the original data.
For example, in the first and third embodiments to be described later, the clone generation processing unit 32 makes a refresh opportunity close to the data stored in the nonvolatile memory 15 based on the evaluation information when the memory card 1 is activated. It is determined whether or not. Then, clone data generation processing is performed for the data determined to have close refresh opportunities.
In the second embodiment, when the memory card 1 is activated, the evaluation information stored for the data unit in which the refresh opportunity is already close is referred to, and clone data generation processing is performed for the data.

The management information processing unit 33 updates the contents (management information) of the management information unit 41. In particular, when clone data is generated by the clone generation processing unit 32, the management information is updated so that either the original data or the clone data becomes an access target.
For example, in the first and second embodiments, the management information processing unit 33 uses the clone data as the main data to be accessed and the sub data from which the original data is not the access target for the data for which the clone data has been generated. Management information is updated so that it is managed as
Further, in this case, when it is determined that the main data has reached the refresh opportunity, the management information processing unit 33 replaces the main / sub data. That is, the management information is updated so that the original data is managed as main data to be accessed.
Further, the management information processing unit 33 updates the management information so that one of the main data and the sub data is discarded for the data for which the clone data is generated at the time of activation.

In the third embodiment, the management information processing unit 33 manages the data for which the clone data is generated as the main data that is the access target of the original data and the sub data that is not the access target of the clone data. Management information is updated.
In this case, when the main data reaches the refresh opportunity, the management information processing unit 33 replaces the main / sub data. That is, the management information is updated so that the clone data is managed as main data to be accessed. Also in this case, the management information processing unit 33 updates the management information so that one of the main data and the sub data is discarded for the data for which the clone data is generated at the time of activation.

The evaluation information processing unit 34 detects or generates reliability evaluation information (value such as the number of bit errors) for the data stored in the nonvolatile memory 15 and accesses the evaluation information table 42. Add / update to.
The evaluation information processing unit 34 also performs processing for clearing the evaluation information about the data for which the clone generation processing unit 32 has generated the clone data from the evaluation information table 42.
In the case of the second embodiment, the evaluation information processing unit 34 determines whether or not the refresh opportunity is close to the data stored in the nonvolatile memory 15 and accesses the refresh opportunity. When it is determined that the data are close to each other, the above-described evaluation information for the data is generated and added to the evaluation information table 42.

The refresh control unit 35 has a function of controlling execution of a refresh operation for data stored in the nonvolatile memory 15. Simply speaking, the refresh operation is an operation of copying the data to another area in the nonvolatile memory 15 in a certain data unit. The operation is substantially the same as the above-described clone data generation. In the case of the present embodiment, the refresh operation rarely occurs, but when refresh is performed, the refresh control unit 35 controls execution of refresh by data copy on the nonvolatile memory 15.

<2. Reduction in processing speed and variation due to refresh>

Here, after describing the processing speed in the case of the previous access leading to the present embodiment, an outline of a technique for eliminating the decrease in the processing speed and the speed variation in the present embodiment will be described.

First, referring to FIG. 3, the case where refresh is performed by the conventional method will be described.
As described above, in recent years, due to the process miniaturization of NAND flash memory, the phenomenon that data bits are garbled easily occurs when the written data is repeatedly read, and it cannot be corrected. Bits are becoming more garbled.
FIG. 3B shows the relationship between the number of reads / writes and the reliability of the semiconductor nonvolatile memory. As shown in the figure, the semiconductor nonvolatile memory has a characteristic that the reliability of data is lowered by repeatedly performing the reading process and the rewriting process.
In a storage system using such a semiconductor non-volatile memory, a method of refreshing data on the non-volatile memory is used as a method for preventing a decrease in data reliability when a value representing the reliability exceeds a certain threshold. It has been.
Specifically, the refresh operation is performed immediately when the number of bits garbled at the time of reading or the like, the number of times of reading or rewriting exceeds a certain threshold.

FIG. 3A shows the operation. FIG. 3 shows a case where the conventional method is executed as the operation of the host device 3, the control unit 11, and the nonvolatile memory 15.
When a read command is transmitted from the host device 2, the control unit 11 executes a read operation for the nonvolatile memory 15. As a result, data is read from the nonvolatile memory 15 to the buffer RAM 14. During this time, the nonvolatile memory 15 is in a busy state.
The control unit 11 performs error detection and error correction on the data read out to the buffer RAM 14, and determines the number of bits garbled as a result of error detection. That is, it is determined whether or not the data read this time needs to be refreshed.
When it is determined that refresh is necessary, the control unit 11 executes a refresh operation. That is, in the nonvolatile memory 15, the current data is copied to another area in the nonvolatile memory 15, and the original data is replaced with copy data. During this time, the nonvolatile memory 15 is in a busy state.
When the refresh is completed, the read data is transferred to the host device 2.

In this way, by using an evaluation index such as the number of bits garbled and refreshing as soon as it exceeds the threshold, the data is replaced before an error that cannot be corrected occurs and the reliability is restored. .
However, when the refresh operation is performed in this way, the processing speed viewed from the host device 2 is reduced. That is, the time from command issue to read data reception becomes longer.

In addition, the processing speed may vary depending on whether the refresh operation occurs or not.
FIG. 4 shows the processing time when the refresh operation is not performed and when it is performed.
It is assumed that the host device 2 transmits a read command at time ts1 and the control unit 11 executes a read operation. In this case, it is not determined that the refresh is necessary, and it is assumed that the read data is transferred to the host device 2 at the time ts2.
On the other hand, it is assumed that the host device 2 also transmits a read command at time ts3 and the control unit 11 executes a read operation. In this case, it is determined that refresh is necessary, a refresh operation is performed, and then read data is transferred to the host device 2 at time ts4.
As in each of these cases, the processing time seen from the host device 2 differs greatly from the times T1 and T2 depending on whether or not the refresh operation is executed.

For these situations, in the present embodiment, while maintaining the reliability of the data in the nonvolatile memory 15, the opportunity to perform the refresh operation is minimized, and the processing speed is prevented from being reduced and the processing time is almost varied. Try not to.
For this reason, in the present embodiment, a value indicating reliability (evaluation information) is used instead of a case where a refresh operation is required, such as when a value indicating the reliability of data exceeds a certain threshold. Data that is likely to require a refresh operation is estimated, and the data is copied as clone data in advance. This prevents a decrease in reliability without reducing the processing speed during reading.

Specific examples will be described later as first to third embodiments. In this example, the processing time is almost as shown in FIG.
In FIG. 5, it is assumed that the host device 2 transmits a read command at time ts11 and the control unit 11 executes a read operation. In this case, it is not determined that refresh is necessary, and it is assumed that the read data is transferred to the host device 2 at time ts12.
On the other hand, it is assumed that the host device 2 also transmits a read command at time ts13 and the control unit 11 executes a read operation. In this case, it is assumed that the situation is determined to require refreshing. In this case, however, the refresh operation is not executed and the management information is updated. This is a process in which one of the clone data and the original data is the data to be accessed in the management information until then, and the other is replaced with the data to be accessed.
This processing is performed, and the read data is transferred to the host device 2 at time ts14.

The management information update can be completed in an extremely short time compared to the refresh operation. As a result, the processing time viewed from the host device 2 hardly changes between times T3 and T4 regardless of the determination of the necessity of refresh.
Even if the necessity of refresh is determined, since the access target is replaced between the clone data and the original data after that, the reliability of the data from the next access is recovered.

<3. First Embodiment>
[3-1: Operation overview]

Hereinafter, the first embodiment will be described as a specific processing example.
In the first embodiment, when the memory card 1 is activated, the refresh operation is similar for each data stored in the nonvolatile memory 15 using the evaluation information recorded based on the value representing the reliability. Guess if it will be needed in the future.
Each data to be determined is described as data in units of physical blocks, for example, but is not limited thereto. Since the erasure unit is a physical block, it is sufficient that the data is at least a unit of the physical block unit. For example, a unit managed as one data file on the management information may be used, or a unit obtained by dividing the data file into several units may be used. It may be a unit of a plurality of physical blocks.

For data determined to require a refresh operation soon, clone data is copied in advance (at startup). That is, the corresponding data is held in the nonvolatile memory 15 in the form of original data and clone data. On the management information, one of clone data and original data is managed as access target data.
In the management information, data managed as an access target, that is, data accessed in response to a command from the host is referred to as “main data”, and the other is referred to as “sub data”.
In the first embodiment, the clone data side is the main data.

Actually, when there is an access request from the host device 2, the control unit 11 accesses the main data based on the management information, and transfers the read data to the host device 2.
In this case, if the read data is in a state that requires immediate refresh, the main data and the sub data are switched. This restores data reliability.

  FIG. 6 schematically shows the transition of such an operation with a certain data DT in the nonvolatile memory 15 as a representative example. As states ST0 to ST6, states relating to the data DT in the nonvolatile memory 15 at each time point are shown. The data DT is, for example, one physical block unit.

  For example, assume that the state is the state ST0 in FIG. That is, it is assumed that data DT in the nonvolatile memory 15 is stored. Here, the data DT is represented as “DTm” and “DTs” in FIG. 6, but “DTm” is managed as an access target according to the read command in the main data, that is, the management information. Means data. On the other hand, “DTs” means secondary data that is not managed as an access target. Of course, the data DT for which no clone data is formed is “DTm”.

  First, at the time of activation, the control unit 11 determines whether or not a refresh operation is necessary for each data unit. When it is determined that the data DTm in the state ST0 is not data that will soon require a refresh operation, it remains as it is. That is, in the state ST1, the data DTm becomes an access target for the subsequent read command.

On the other hand, if it is determined that the refresh opportunity is close for data DTm, clone data is generated as shown in state ST2.
Further, in the state ST3, the management information is updated so that the generated clone data is the main data and the original data is the sub data.
Therefore, for subsequent read commands, the data DTm, which is clone data, becomes an access target. The clone data is fresh data at this point, and reliability is restored compared to the original data. Therefore, even if there is a subsequent read opportunity, it is rarely determined that an immediate refresh is necessary until a considerable number of reads.

After the normal operation of the memory card 1 is continuously performed in this state ST3, the operation of the memory card 1 is turned off once the connection from the host device 2 is canceled or the host device 2 is turned off. Suppose that After that, it is assumed that the memory card 1 is activated in connection with connection with the host device 2 or power-on of the host device 2.
At the time of starting in that case, the state is ST5. In this case, one of the main data and the sub data, for example, the sub data DTs is discarded (erased). Then, the process returns to the state ST0 and the same processing as described above is performed. The reason why the sub-data (original data side) is discarded is that the original data is judged to be close to the refresh opportunity at the time of the previous activation, and the reliability is relatively lowered. That is, it is considered that it is preferable to leave the main data (clone data side) in terms of reliability.
Note that the clone data side may be accessed very many times during the previous activation, resulting in a decrease in reliability and close refresh opportunities. However, in this case, since the clone data is newly generated with the previous clone data as the original in the state ST0 → ST2 at the time of this activation, the reliability is restored.

By the way, when the normal operation of the memory card 1 is continuously performed in the state ST3 as described above, it may be determined that an immediate refresh is necessary when reading the data DTm at a certain time. .
In this case, as shown as state ST4, the management information is updated so that the main data and the sub data are exchanged. In this case, the management information is updated so that the original data is the main data where the clone data is the main data.
Therefore, when a subsequent read opportunity occurs, the original data side is accessed.

The original data side is data determined to be close to a refresh opportunity at the time of activation, but this is not a level that requires immediate refresh. Accordingly, normal data reading is still possible.
Therefore, in this case, even if the reliability is deteriorated even when the clone data side needs to be refreshed, it is accompanied by the original data side, thereby eliminating the necessity of performing an immediate refresh at the time of reading.
In the first place, since the clone data side, which is fresh data, is the main data in the state ST3, it is rare that the clone data side is determined to be immediately refreshed during the continuous activation. This is a case where a large number of reads are concentrated on the same data during startup. Even in such a rare case, if the data DTm (clone data side) is required to be refreshed immediately, the use of the original data side can eliminate the need for an immediate refresh operation. .

If the normal operation of the memory card 1 is continued in this state ST4 and then turned off, and then the memory card 1 is activated, the state ST6 is entered. Also in this case, one of the main data and the sub data, for example, the sub data DTs is discarded (erased). Then, the process returns to the state ST0 and the same processing as described above is performed.
The reason why the sub data (in this case, the clone data side) is discarded is that the clone data has already been determined to be immediately refreshed, and reliability has been lowered. In other words, it is because it is still considered preferable in terms of reliability to leave the main data (original data side).
In this case, since the original data side is determined to be close to the refresh opportunity at the previous activation, the reliability is not so high. However, since the clone data is newly generated with the original data side as the original again in the state ST0 → ST2 at the time of this activation, the reliability is restored.

When the normal operation of the memory card 1 is continuously performed in the above-described state ST4, it is determined that an immediate refresh is necessary when reading the data DTm (in this case, the original data side) at a certain point in time. Sometimes.
In that case, a refresh operation is performed on the data DTm in order to deal with a decrease in reliability.
As described above, it is rare that the state ST4 is originally reached (the clone data side needs immediate refresh). In addition, it is rare that the operation is continued in the state ST4 and the original data side needs to be immediately refreshed. Therefore, it can be said that there is practically no opportunity to execute the refresh operation. However, since it cannot be said that there is nothing, a refresh operation is prepared as compensation in that case.

It may be considered that after the refresh operation is performed, the state returns to the state ST4. However, the reliability of the data DTm in that case is recovered.
After that, when the operation of the memory card 1 is turned off and the memory card 1 is further activated, the state is ST6. Also in this case, one of the main data and the sub data, for example, the sub data DTs is discarded (erased). Then, the process returns to the state ST0 and the same processing as described above is performed.
The reason why the sub data (in this case, the clone data side) is discarded is that the clone data has already been determined to be immediately refreshed, and reliability has been lowered. On the other hand, the main data in this case is refresh data, and the reliability has been recovered.
Note that the refresh data side may be accessed very many times after the refresh during the previous activation, so that the reliability is lowered and the refresh opportunity is close. However, in this case, the clone data is newly generated with the previous refresh data as the original in the state ST0 → ST2 at the time of the current activation, so that the reliability is recovered.

In the first embodiment, the data DT changes as described above, thereby maintaining the reliability and reducing the access time and stabilizing by minimizing the refresh operation.

[3-2: Information for evaluation]

Here, the evaluation information for determining the proximity of the refresh opportunity performed by the control unit 11 at startup will be described.
The data DT (DTm, DTs) illustrated in FIG. 6 is, for example, data in units of physical blocks. In that case, in order to determine the proximity of the refresh opportunity for the data stored in the nonvolatile memory 15, the evaluation information only needs to be information that can evaluate the reliability for each physical block.

A physical block as an erasure unit is composed of a plurality of physical pages PU as shown in FIG. 7A (physical pages with page numbers 0 to n).
As evaluation information for one physical block, various values may be stored for the entire physical block, or various values may be stored for each physical page.

FIG. 7B shows an example of evaluation information corresponding to one certain physical block PBx.
In this example, for each physical page, the emergency level, the previous emergency level, the previous emergency level, the cumulative number of emergency levels, the maximum number of emergency levels, the number of reads, the number of erases (rewrites), and the evaluation information It is supposed to be.
Further, the cumulative number of emergency levels, the maximum number of emergency levels, the number of reads, and the number of erasures (rewrites) as a whole of the physical block PBx are also stored.

As the emergency level, for example, the number of bit errors (bit corruption) at the time of reading may be used as it is. Alternatively, the number of bit errors may be divided into a plurality of ranges to form a corresponding emergency level value. In any case, the emergency level may be information that can be used to estimate the necessity of refresh.
As the emergency level, for example, as the number of bit errors at the time of reading in each physical page, the current (at the time of reading) and the previous and previous times are stored. Of course, you may go back and memorize at the time of the past lead.
The cumulative number of bit errors and the maximum number so far can also be used for reliability estimation.
In addition, the number of reads and the number of erases for each physical page can also be used for reliability estimation.

FIG. 7B is merely an example of the content of the evaluation information.
For example, since it is only necessary to evaluate the reliability of data in units of physical blocks, it is not necessary to include information for each physical page in the evaluation information. That is, only the cumulative number of emergency levels, the maximum number of emergency levels, the number of reads, and the number of erases (rewrites) for the entire physical block PBx at the right end in FIG. 7B may be used as the contents of the evaluation information.
Alternatively, if the information for each physical page is stored, the value of each information can be calculated in units of physical blocks, so only the information for each physical page may be used as the content of the evaluation information.
Further, for example, the number of bit errors of only one (or a plurality of) physical pages such as a physical page of page number 0 may be stored and handled as an evaluation value representing a physical block.
Further, the information content included in the evaluation information may be considered other than the above. Only the emergency level or the number of reads may be used. Further, only the number of erasures may be used.

  Here, assuming that the above-described clone data generation is performed in units of physical blocks, it is necessary to have the evaluation information in units of physical blocks. For example, clone data generation may be performed in units of multiple physical blocks, data files, or the like. It is also possible to do. In that case, the evaluation information may be provided in units of the plurality of physical blocks or in units of data files. In those cases, the emergency level of each physical block may be stored, or the emergency levels of some representative physical blocks may be stored. Of course, information such as an emergency level in a physical page may be stored in that case.

In the first embodiment, each value of such evaluation information is added or updated and stored at the time of reading or writing (erasing / rewriting).
The information content for the physical block PBx or physical page PU may be updated every time the physical block PBx or physical page PU is read or written. You may make it update to. The emergency level may be stored only when the read count is divisible by the predetermined value m, or the read count or erase (the rewrite / write count is a predetermined count (for example, the first time, the tenth time, the 50th time, the 100th time) In other cases, the emergency level may be stored when the condition is met, etc. The evaluation information may be updated and recorded only when certain conditions are met.

Further, the number of pieces of information to be held is not limited to one per physical page unit or physical block unit.
For example, when the current emergency level is stored as evaluation information at the time point k0 for one physical block PBx, the evaluation information for the physical block PBx is additionally stored at each time point k1, k2,. It is possible to keep it.

  The evaluation information may be reset (cleared) at the time of activation, when clone data is generated, or when the corresponding data unit (physical block or the like) is rewritten (erased). However, when information on the number of times of rewriting (erasing) is included, it is appropriate that the value of the number of times of rewriting should not be reset (cleared) in principle.

  Such evaluation information may be recorded on the nonvolatile memory 15 every time the information is updated, or temporarily recorded in the SRAM 12 or the like in the system, and recorded in the nonvolatile memory 15 at a fixed timing. You may make it do.

When determining the proximity of a refresh opportunity at startup, whether each physical block needs to be refreshed in the near future based on the emergency level, read count, erase count, etc. for each physical block stored at that time Information for estimating whether or not (reliability information) is calculated. Alternatively, each stored information may be used as reliability information as it is.

[3-3: Processing example]

As the first embodiment, a process of the control unit 11 for realizing the operation described in FIG. 6 will be described. The process described below is executed by the control unit 11 having the functions described in FIG.

  FIG. 8 shows processing executed by the control unit 11 when the memory card 1 is activated. 2, step F101 is a function of the management information processing unit 33, steps F102 to F104 are functions of the clone generation processing unit 32, and step F105 is a process of the management information processing unit 33 and the evaluation information processing unit 34. Performed by function.

In step F101, the control unit 11 performs processing for discarding one of the main and sub data. For each data unit stored in the non-volatile memory 15, clone data is generated in the past, and there are main data and sub data in the management information. Therefore, at the time of activation, for the purpose of once canceling this state, a data unit in which main / sub data is present is retrieved from the management information, and one of the main / sub data is discarded for the corresponding data unit.
Specifically, the management information may be updated so as to cancel the management state of one of the main data and the sub data, and it is not actually necessary to erase data on the nonvolatile memory 15 at this time.
In the case of the first embodiment, as described in the states ST5 and ST6 in FIG. 6, the sub data is discarded.
By the process of step F101, all data stored in the nonvolatile memory 15 is in a state ST0 in FIG. 6 at the time of activation.

  In step F102, the control unit 11 analyzes the evaluation information for each data. For example, for each physical block in which data is stored, the evaluation information is analyzed, and it is determined whether or not the reliability of the physical block is lowered to the extent that it will become a refresh opportunity in the near future. . A specific example of the determination will be described later.

  If it is determined that all the data are not in the vicinity of the refresh opportunity, the processing of FIG. This is a case where all data is in the state ST1 in FIG.

When it is determined that at least one or more pieces of data are close to the refresh opportunity, the control unit 11 proceeds from step F103 to F104 to generate clone data. That is, for the data determined to be close to the refresh opportunity, copy data is generated in another area in the nonvolatile memory 15. The data is in the state ST2 in FIG.
Further, the control unit 11 updates management information and clears unnecessary evaluation information in step F105.
The management information is updated so that the generated clone data is the main data to be accessed and the original data is the sub data. That is, state ST3 in FIG.
As the information for evaluation, the data for which the clone data has been generated is the information for evaluation of the physical block of the original data at that time, and since the clone data is accessed after this time, it is cleared as unnecessary. .
However, in FIG. 6, the original data side may be accessed in state ST4, and then the state ST0 may be reached, and processing that is not cleared is necessary as necessary. (Additionally, even in this case, since the evaluation information is newly generated when the original data side is actually accessed, it is not always necessary to clear and store the information.)

As described above, when clone data is generated, data determined to be close to the refresh opportunity (for example, physical block unit) is in the state ST3 in FIG. 6, and other data is in the state ST1.
The control unit 11 waits for a command from the host device 2 after performing the above processing at startup.

FIG. 9 shows processing of the control unit 11 when a read command is transmitted from the host device 2.
2, steps F151 and F158 are functions of the access control unit 31, steps F152 are functions of the information processing unit for evaluation 34, steps F153 → F154 → F156 are functions of the refresh control unit 35, and steps. F153 → F154 → F155 is executed as a function of the management information processing unit 33, and step F157 is executed as a function of the management information processing unit 33 and the evaluation information processing unit 34.

  When there is a read command from the host device 2, the control unit 11 executes a read operation to the nonvolatile memory 15 in step F151. That is, the data specified by the logical add in the read command is converted into a physical address with reference to the management information, and the data of the physical add in the nonvolatile memory 15 is read and transferred to the buffer RAM 14. Further, the control unit 11 performs error detection / error correction processing on the data fetched into the buffer RAM 14.

  In step F152, the control unit 11 updates the evaluation information. For example, the number of bit errors in the current read is added to the evaluation information, and the evaluation information is updated so that the number of reads is incremented by one.

Here, in step F153, the control unit 11 determines the reliability of the current read data, for example, whether the number of bit errors exceeds a predetermined threshold or whether the number of reads exceeds a threshold. Do. Here, the reliability determination is a determination as to whether or not the reliability has decreased to the extent that immediate refresh is necessary.
If it is determined that the reliability is maintained without satisfying these conditions, the process proceeds to step F158 and the read data is transmitted to the host device 2. That is, the data that has been taken into the buffer RAM 14 and error-corrected is transmitted to the host device 2 via the device interface 13.

  On the other hand, if it is determined at step F153 that the reliability is reduced, that is, it is necessary to refresh to maintain the reliability in the future, the control unit 11 reads the current main data, that is, this time, at step F154. The processing branches depending on whether the data is generated as clone data at the time of activation or is original data.

At step F154, the main data is clone data in the case of state ST3 in FIG. In this case, the control unit 11 proceeds to step F155 and updates the management information. In this case, the management information is updated so that the main data and the sub data are exchanged. Therefore, the original data side at the time of activation is set as main data, and the clone data side is set as sub data. That is, state ST4 in FIG.
In step F158, the read data is transferred and the processing corresponding to the access command is completed.
In this case, when the next read command is generated, the original data side at the time of activation is read as main data.

In step F154, the main data is not clone data in the case of state ST4 or state ST1 in FIG.
Here, the case where the state is ST4 is a case where after the clone data is required to be refreshed, the original data side is also required to be refreshed.
Therefore, the control unit 11 causes a data refresh operation to be executed in step F156.
With the refresh operation, the management information and the evaluation information are updated in step F157. The management information is updated so that the new data copied by the refresh operation becomes the main data. With respect to the evaluation information, since the previous evaluation information is no longer necessary by the current refresh operation, it is cleared.
In step F158, the read data is transferred and the processing corresponding to the access command is completed.

  Even when it is determined in step F154 that the state is ST1 and the main data is not clone data, the refresh operation, the management information update, and the evaluation information are cleared. However, since the state ST1 is data that has been determined that the refresh opportunity is still far from the start-up, such a situation is rare.

  The operation described with reference to FIG. 6 is realized by the control unit 11 performing the processes of FIGS.

Here, an example of the refresh opportunity proximity determination process using the evaluation information in step F102 of FIG. 8 will be described.
In this example, the evaluation reference value including the number of bit error occurrences, the number of reads, and the number of rewrites (erase) is used to derive a criterion value used to estimate whether or not a refresh operation is necessary soon. For example.
In the first example and the second example, a linear equation is obtained based on the number of recorded bit errors and the number of reads, and the number of times of reading until the limit at which reading from the nonvolatile memory 15 becomes impossible is calculated. Data (physical block or the like) with a small number of remaining readings until the limit is reached is determined as a refresh opportunity proximity.
In the third example, the number of bit error occurrences and the number of reads are directly used as judgment reference values.

First example For each data (for example, for each physical block), a bit error occurs when the number of read times (x) in the data is the number of read times (x) and the number of read times (x + n) thereafter A linear equation is derived from the two values of the cumulative number (y + m) of the number. At this time, the interval (n) between the two points is determined based on the cumulative number of rewrite (erase) times. However, any interval may be used. For example, when n = 50, the values are, for example, the 50th and 100th reads. The actual determination depends on the number of read times at which the evaluation information is stored, but for example, the most recent read time point and the value of the read number time n times before that are used. That's fine.

In FIG. 10A, a linear equation on the coordinates of the number of reads and the emergency level is indicated by a broken line. In this case, the emergency level corresponds to the cumulative number of bit errors.
Using the derived linear equation, the number of readings reaching the limit at which reading from the nonvolatile memory 15 becomes impossible is calculated. For example, if the bit error accumulation number E1 is a reliability limit, the limit of the number of reads is N1 from the linear equation.
Here, the remaining number of reads until reaching the limit is N1-N2 NR times. By comparing the NR times with a certain threshold and determining whether or not the remaining number is small, it is possible to determine whether or not the refresh opportunity is close.

Second Example For each data, each time the number of reads reaches a certain number, an approximate expression of a linear equation is calculated using the least square method based on the maximum number of bit error occurrences at that time. The linear equation calculated | required by the approximate expression is shown with the broken line in FIG. 10B.
Using the derived linear equation (approximate equation), the number of readings reaching the limit at which the non-volatile memory 15 cannot be read is calculated in the same manner as in the first example. For example, the limit number N1 of the number of reads is obtained. The remaining number of reads until reaching the limit is obtained as N1-N2 NR times, and this NR number is compared with a certain threshold value to determine whether the remaining number is small. This makes it possible to determine whether or not the refresh opportunity is close.

  In either case of the first example or the second example, the accumulated number of rewrites (erasures) may be considered as a coefficient in the slope of the linear equation as necessary.

Third Example The maximum number of bit error occurrences in each data (for example, each physical block) is used as a determination reference value, and the upper fixed number of data (for example, physical blocks) having the largest maximum number is determined to be near the refresh opportunity. That is, a physical block having a relatively large number of bit error occurrences is extracted from each physical block, and is set as a physical block near the refresh opportunity.
When there are a plurality of physical blocks having the same maximum number, the number of reads in the physical block is also used as a determination reference value, and the higher number of reads is given priority.
If the maximum number of bit errors is small but the number of reads in a physical block is greater than or equal to a certain reference value, it may be determined as a physical block close to a refresh opportunity as another frame.

Although the first embodiment has been described above, the following effects can be obtained by the first embodiment.
In the conventional method, when a refresh operation is required, the refresh operation is immediately performed, which causes a decrease in performance during reading. Further, since the refresh operation is started by an internal factor such as whether or not the number of occurrences of bit errors in the target data exceeds a certain threshold value, the read performance varies.
On the other hand, in the method of the present embodiment, first, in the case of state ST3 in FIG. 6, the main data (clone data side) can be repeatedly read up to the number of times of reading limit. That is, it is possible to continue to read until the number of times of reading N1 that becomes the reliability limit E1 in FIG. Further, even when refreshing is necessary due to the repeated data being read and the data becomes garbled, the process moves to the state ST4 and the main data (original data side) can be repeatedly read up to the number of times of reading limit.
Eventually, the opportunity to perform a refresh operation at the time of reading while maintaining the reliability of data can be almost eliminated, and there is almost no decrease in performance (reading processing time) and variations at the time of reading.

Further, by managing the clone data replicated at the time of activation as the main data and using the original data side as the secondary data, the evaluation information can be reset (cleared) each time the clone data is replicated. For this reason, at the next start-up, there is a lower possibility that a refresh operation will be necessary.
For this reason, unless the number of readings to specific data within one start-up is extremely large, each time start-up requires clone data from data different from the original data at the previous start-up (previous clone data) If so, it is duplicated. Therefore, under a use situation where the activation is repeated, there is no data that needs a refresh operation soon, and it is possible to maintain a highly reliable state as data.

In addition, in the determination of the proximity of refresh opportunities as a precondition for replicating clone data, in the first example and the second example described above, data for which the remaining number of reads until reaching the limit is less than a certain reference number is refreshed. Proximity. As a result, when there is no target data that needs to be refreshed in the near future, it is possible to avoid duplicating clone data using the storage area of the nonvolatile memory 15 in vain. This is preferable because the frequency of use of each area of the nonvolatile memory 15 is not increased more than necessary.
The same applies to a case where the maximum number of bit error occurrences equal to or greater than a certain reference number is determined as a refresh opportunity proximity and is targeted for clone data generation as in the third example.

<4. Second Embodiment>

Next, a second embodiment will be described.
In the second embodiment, when the control unit 11 (the evaluation information processing unit 34) accesses the data stored in the nonvolatile memory 15, the control unit 11 determines whether or not the refresh opportunity is close to the data. When it is determined that refresh opportunities are close, evaluation information for the data is generated. Then, the control unit 11 (clone generation processing unit 32) performs clone data generation processing on data that is determined to be close to the refresh opportunity based on the evaluation information at the time of activation.

For example, as shown in FIG. 12A, x storage frames are set as the evaluation information table 42. Here, when the evaluation information is stored in units of physical blocks, the physical block number and the number of bit errors generated when reading the physical block are stored in each storage frame as evaluation information as necessary. To do.
For example, when a read is performed on a physical block and the number of bit errors occurring in data read from the physical block is greater than or equal to a predetermined number, it is determined that the refresh opportunity is close and the number of the physical block (physical Block address) and the number of bit errors are added to the evaluation information table 42.

FIG. 12 shows an example in which when the number of bit errors is 4 or more, it is determined that the refresh opportunity is close and the evaluation information table 42 is updated.
FIG. 12A shows an example of the evaluation information stored in the evaluation information table 42 at that time when the number of bit errors in the physical blocks PB105, PB8, PB35, and PB36 is 4 or more in the access performed after activation. The contents are shown.
Thereafter, data is read in response to a read command from the host device 2. For example, if the number of bit errors is 6 when reading the physical block PB210, the information is for evaluation as indicated by the hatched portion in FIG. 12B. It is added to the information table 42. Similarly, information on physical blocks with a large number of bit errors is added as evaluation information.

FIG. 12C shows a state in which the evaluation information in the shaded area is further added and all the x storage frames in the evaluation information table 42 are filled.
For example, if a physical block with a large number of bit errors is discovered by accessing after such a state, for example, the contents of the physical block with a small number of bit errors in the stored evaluation information may be rewritten. .
For example, if the number of bit errors is 6 when the physical block PB511 is newly accessed, the physical block PB36 having the smallest number of bit errors of 4 is deleted from the evaluation information table at the time of FIG. As in the section, the evaluation information of the physical block PB511 is stored.

By doing so, the evaluation information table 42 indicates x or less physical blocks estimated to be close to the refresh opportunity.
At the time of activation, the control unit 11 may automatically determine that the physical block listed in the evaluation information table 42 is a physical block near the refresh opportunity.
Here, the evaluation information is stored in units of physical blocks. However, as a unit of “data” for generating the evaluation information, a data file unit, a plurality of physical block units, or the like may be used.

The processing of the second embodiment will be described with reference to FIGS.
In the second embodiment, the transition of the data state is the same as that of the first embodiment described with reference to FIG. 6, and will be described with reference to FIG.

  FIG. 13 shows processing executed by the control unit 11 when the memory card 1 is activated. 2, step F201 is a function of the management information processing unit 33, steps F202 and F203 are functions of the clone generation processing unit 32, and step F204 is a process of the management information processing unit 33 and the evaluation information processing unit 34. Performed by function.

In step F201, the control unit 11 performs processing for discarding one of the main and sub data. As in the case of the first embodiment, the sub data is discarded as described in the states ST5 and ST6 of FIG.
By the process of step F201, all data stored in the nonvolatile memory 15 is in the state ST0 in FIG. 6 at the time of activation.

In step F202, the control unit 11 branches the process depending on the necessity of generating clone data. In the case of the second embodiment, as described above, if there is data (such as a physical block unit) listed in the evaluation information table 42, it is regarded as data near the refresh opportunity. That is, in step F202, the control unit 11 determines whether or not one or more data is stored in the evaluation information table 42.
If no data is listed in the evaluation information table 42, the processing of FIG. 13 is terminated from step F202. This is a case where all data is in the state ST1 in FIG.

If at least one or more data is listed in the evaluation information table 42, the control unit 11 proceeds from step F202 to F203, and generates clone data. That is, copy data is generated in another area in the nonvolatile memory 15 for the data that has been determined to be close to the refresh opportunity. The data is in the state ST2 in FIG.
In step F204, the control unit 11 updates management information and clears unnecessary evaluation information. This is the same processing as step F105 in FIG. 8 of the first embodiment.
As a result, the data (for example, physical block unit) for which clone data has been generated is in the state ST3 in FIG. 6, and the other data is in the state ST1.
The control unit 11 waits for a command from the host device 2 after performing the above processing at startup.

FIG. 14 shows processing of the control unit 11 when a read command is transmitted from the host device 2.
2, steps F251 and F259 are functions of the access control unit 31, steps F252 → F255 → F257 are functions of the refresh control unit 35, and steps F252 → F255 → F256 are functions of the management information processing unit 33. Steps F253 and F254 are executed as functions of the evaluation information processing unit 34, and step F258 is executed as functions of the management information processing unit 33 and the evaluation information processing unit 34.

  When there is a read command from the host device 2, the control unit 11 executes a read operation to the nonvolatile memory 15 in step F251. That is, the data specified by the logical add in the read command is converted into a physical address with reference to the management information, and the data of the physical add in the nonvolatile memory 15 is read and transferred to the buffer RAM 14. Further, the control unit 11 performs error detection / error correction processing on the data fetched into the buffer RAM 14.

  Here, in step F252, the control unit 11 determines the reliability of the current read data, for example, whether the number of bit errors is a numerical value exceeding a predetermined threshold or whether the number of reads exceeds a threshold. Do. That is, it is a determination as to whether or not the reliability has decreased to such an extent that an immediate refresh is necessary.

If it is determined that the reliability is maintained without satisfying these conditions, the process proceeds to step F253. Here, the control unit 11 determines whether or not the reliability of the access target data has decreased to the extent that it will become a refresh opportunity in the near future.
In this case, for example, it is conceivable to determine whether or not the reliability tends to be lowered under conditions looser than the reliability determination in step F252.
For example, when the number of bit errors is p times as the first threshold value and used for the reliability determination in step F252, q times (where q <p) is the second threshold value for determining the proximity of the refresh opportunity. And
The number of reads and the number of rewrites are also managed, and these may be used for the refresh opportunity proximity determination under milder conditions than the reliability determination of whether or not immediate refresh is necessary.

  If it is determined that the access target data is not near the refresh opportunity, the control unit 11 proceeds to step F259 as it is and transmits read data to the host device 2. That is, the data that has been taken into the buffer RAM 14 and error-corrected is transmitted to the host device 2 via the device interface 13.

  On the other hand, when it is determined in step F253 that the refresh opportunity is close, the control unit 11 updates the evaluation information table 42 so as to add the evaluation information related to the data to the evaluation information table 42. That is, the processing described with reference to FIG. The data added in this step F254 and holding the evaluation information in the evaluation information table 42 thereafter becomes the clone data generation target data in step F203 in FIG.

  On the other hand, if it is determined at step F252 that the reliability is lowered, that is, it is necessary to refresh to maintain the reliability in the future, the control unit 11 reads the current main data, that is, this time, at step F255. The processing branches depending on whether the data is generated as clone data at the time of activation or is original data.

At step F255, the main data is clone data in the case of state ST3 in FIG. In this case, the control unit 11 proceeds to step F256 and updates the management information. In this case, the management information is updated so that the main data and the sub data are exchanged. Therefore, the original data side at the time of activation is set as main data, and the clone data side is set as sub data. That is, state ST4 in FIG.
In step F259, the read data is transferred and the processing corresponding to the access command is completed. In this case, when the next read command is generated, the original data side at the time of activation is read as main data.

In step F255, the main data is not clone data in the case of state ST4 or state ST1 in FIG.
Here, the case where the state is ST4 is a case where after the clone data is required to be refreshed, the original data side is also required to be refreshed.
Therefore, the control unit 11 executes a data refresh operation in step F257.
With the refresh operation, the management information and the evaluation information are updated in step F258. The management information is updated so that the new data copied by the refresh operation becomes the main data. With respect to the evaluation information, since the previous evaluation information is no longer necessary by the current refresh operation, it is cleared.
In step F259, the read data is transferred and the processing corresponding to the access command is completed.
Even when it is determined that the main data is not clone data in the state ST1 at the time of step F255, the refresh operation, the management information update, and the evaluation information are cleared.

  The second embodiment in which the control unit 11 performs the processes of FIGS. 13 and 14 described above, and the operation described in FIG. 6 is performed, and the determination of the proximity of the refresh opportunity is performed when the evaluation information is additionally updated. Will be realized.

In the second embodiment as described above, the following effects are obtained in addition to the effects of the first embodiment. That is. In the first embodiment, an example in which a determination reference value is derived based on evaluation information at the time of startup and a refresh opportunity proximity determination is performed has been described. In the second embodiment, such a determination is performed at the time of startup. There is no need to do this for all data. Therefore, the processing load on the control unit 11 at the time of activation is reduced.
Further, when there is no data to be judged (estimated) that will be required in the near future for the refresh operation, no evaluation information is recorded in the evaluation information table 42. Therefore, the recording amount (size) of the evaluation information can be reduced.

In FIG. 12, the number of storage frames for evaluation information stored in the evaluation information table 42 is x. However, no particular limitation is imposed, and all data determined to be close to the refresh opportunity in step F253 (capacitatively). Information for evaluation may be stored (as much as possible).
As the contents of the evaluation information, not only the number of bit errors corresponding to the physical block as shown in FIG. 12, but also the number of reads and the number of rewrites may be used, or these may be used in combination.
Furthermore, if all the data (physical blocks and the like) listed in the evaluation information table 42 are to be generated as clone data as a result, the evaluation information table 42 contains data determined to be close to the refresh opportunity at the time of reading. Only the identification information (for example, physical block address) may be stored.

It is also conceivable to add information for evaluation of data to be accessed to the evaluation information table 42 every time access is executed. In this case, however, x storage frames are set in the evaluation information table 42 as shown in FIG. 12, and the update is performed so that, for example, the one with a large number of bit errors remains as described in FIGS. 12C to 12D. I will do it. Of course, the number of reads and the number of rewrites may be used.
Then, in the evaluation information table 42, x data having a large number of bit errors remain among the data to be accessed, and as a result, the determination of the proximity of the refresh opportunity is made at the time of access. It will be. In this case as well, at the time of activation, the data (physical block) listed in the evaluation information table 42 may be the clone data generation target.

Of course, in that case, among the data remaining in the evaluation information table 42, data having a predetermined number of bit errors or more is set as a clone data generation target, or the upper predetermined number of data is set as a clone data generation target. A part of the evaluation information table 42 may be a clone data generation target.

<5. Third Embodiment>

A third embodiment will be described with reference to FIG. FIG. 15 shows the transition of data as in FIG.
In the third embodiment, for the data for which clone data is generated at the time of startup, the original data is initially managed as main data to be accessed, and the clone data is managed as sub data that is not to be accessed. This is an example.
When the main data (original data side) reaches a refresh opportunity, the management information is updated so that the clone data is managed as the main data to be accessed.

  The start-up time is set to state ST10. At the time of startup, the control unit 11 determines whether or not a refresh operation is necessary for each data unit. When it is determined that the data DTm in the state ST0 is not data that will soon require a refresh operation, the data DTm becomes an access target for the subsequent read command as it is as the state ST11.

On the other hand, if it is determined that the refresh opportunity is close for data DTm, clone data is generated as shown in state ST12.
In the third embodiment, the generated clone data is managed as sub-data, and the main data remains the original data.
Therefore, for subsequent read commands, the data DTm, which is the original data, becomes the access target.
The original data side is data determined to be close to a refresh opportunity at the time of activation, but this is not a level that requires immediate refresh. Accordingly, normal data reading is still possible.

It is assumed that after the normal operation of the memory card 1 is continuously performed in this state ST12, the operation of the memory card 1 is once turned off and then activated.
At the time of starting in that case, the state is ST14. In this case, one of the main data and the sub data is discarded.
In this case, it is preferable to update the management information so that the main data DTm (original data side) is discarded, and at the same time, the sub data DTs (clone data side) is changed to the main data DTm. This is because in this case, the main data DTm (original data side) is judged to be close to a refresh opportunity at the time of the previous activation and is relatively unreliable, while the sub data DTs (clone data side) is fresh data. It is. Therefore, it is considered that it is preferable to leave the clone data side in terms of reliability and also to keep a refresh opportunity away from the data. In other words, the update of the management information in this case is equivalent to the state where the refresh is performed.

  In this case, it is not unthinkable to discard the sub data DTs (clone data side). In this case, at the time of this activation, the main data DTm (original data side) becomes a clone data generation target as a refresh opportunity proximity.

By the way, when the normal operation of the memory card 1 is continuously performed in the state ST12 as described above, it is determined that immediate refresh is necessary when reading the data DTm (original data side) at a certain time. May be.
In this case, as shown as state ST13, the management information is updated so that the main data and the sub data are exchanged. In this case, the management information is updated so that the original data is the main data and the clone data is the main data.
Therefore, when a subsequent read opportunity occurs, the clone data side is accessed.
In other words, in this case, when the reliability of the original data has deteriorated to a state where it is necessary to be fresh, the necessity of performing an immediate refresh is eliminated by replacing the read target with the clone data.

If the normal operation of the memory card 1 is continuously performed in this state ST13, the memory card 1 is turned off and then the memory card 1 is activated, then the state ST15 is entered. Also in this case, one of the main data and the sub data is discarded (erased). Then, the process returns to the state ST0 and the same processing as described above is performed.
In this case, it is appropriate to discard the sub data (original data side) at this time. This is because the original data side is data determined to be refreshed during the previous activation.

When the normal operation of the memory card 1 is continuously performed in the above-described state ST13, it is determined that an immediate refresh is necessary when reading the data DTm (in this case, the clone data side) at a certain point in time. Sometimes.
In that case, a refresh operation is performed on the data DTm in order to deal with a decrease in reliability.
In the state ST13, since the fresh clone data side is the main data DTm, it is rare that an immediate refresh is necessary in the state ST13. Therefore, it can be said that there is practically no opportunity to execute the refresh operation. However, since there is nothing at all, a refresh operation is prepared as compensation in that case.

After the refresh operation is performed, it can be considered that the process returns to the state ST13. However, the reliability of the data DTm in that case is recovered.
Thereafter, when the operation of the memory card 1 is turned off and the memory card 1 is further activated, the state is ST15. Also in this case, it is appropriate to discard (erase) the sub data DTs. This is because the sub data DTs in this case is data on the original data side determined to be refreshed during the previous activation.
Then, the process returns to the state ST0 and the same processing as described above is performed.

In the third embodiment, the data DT changes as described above, thereby maintaining the reliability and reducing the access time and stabilizing by minimizing the refresh operation.
The processing of the control unit 11 for the third embodiment is basically as shown in FIGS. 8 and 9, but the following points may be different as detailed points.

First, in step F101 in FIG. 8, the main data DTm is discarded in the case of state ST14 in FIG. 15, and the sub data DTs is discarded in the case of state ST15.
In step F105, the management information may be updated so that the generated clone data is managed as the sub data DTs.
The evaluation information is as follows.
When the state is changed from ST14 to ST10, the main data DTm is clone data created at the previous activation, and the clone data has not been read yet. Therefore, the information for evaluation up to that point may be cleared for the data.
When the state transitions from ST15 to ST10 and no refresh is performed, the main data DTm is clone data created at the previous activation. However, the clone data may have been read. Therefore, it is preferable to maintain the previous evaluation information for the data.

In step F155 of FIG. 9, the process transits from the state ST12 to ST13 of FIG. 15. Therefore, the management information update is a process of setting the original data side as the sub data DTs and the clone data side as the main data DTm. do it.

<6. Modification>

Although the embodiments have been described above, the following modifications may be considered for the technology of the present disclosure.
In the third embodiment, as in the first embodiment, the proximity of refresh opportunities is determined for each data at the time of startup. However, as in the second embodiment, the data at the time of reading is determined. It is also possible to apply a process for determining the proximity of the refresh opportunity and storing the evaluation information.

The content of the evaluation information is the number of bit errors, the number of reads, the number of rewrites, etc., but other examples are also conceivable.
In addition, after giving different weights to the number of bit errors, the number of reads, the number of rewrites, and the like, an index for determining the proximity of refresh opportunities may be obtained from these data.

  In the embodiment, the clone data generation is performed at the time of activation, but it may be performed at other times than the activation. For example, it can be performed when the operation is turned off or when the memory card 1 is removed.

Moreover, although the example of the memory card 1 has been described in the embodiment, the technology of the present disclosure can be applied even when the nonvolatile memory 15 and the control unit 11 are configured separately. For example, it can be realized as a control device for controlling read / write access to a nonvolatile memory.
The technology of the present disclosure can be applied to various memory cards, SSDs, eMMCs, and the like.

In addition, this technique can also take the following structures.
(1) Data stored in a non-volatile memory, for which it is determined that a refresh opportunity is close based on information for evaluation regarding refresh necessity, is the same as the original data, using the data as original data Clone generation processing unit for performing processing to generate the clone data in the nonvolatile memory,
A management information processing unit that updates management information so that either one of the original data or the clone data becomes an access target at the time of access execution;
An evaluation information processing unit that generates the evaluation information for the data when accessing the data stored in the nonvolatile memory;
A control device comprising:
(2) The management information processing unit manages the data for which the clone data is generated so that the clone data is managed as main data to be accessed and the original data is managed as sub data that is not to be accessed. The control device according to (1), wherein the management information is updated.
(3) When it is determined that the main data has reached a refresh opportunity, the management information processing unit updates the management information so that the original data is managed as main data to be accessed (2) The control device described in 1.
(4) The above-described (3) further comprising a refresh control unit that performs a refresh operation control on the original data when it is determined that the main data has reached a refresh opportunity after the original data is the main data. ).
(5) The management information processing unit stores the management information so that one of the main data and the sub data is discarded for the data for which the clone data is generated when the nonvolatile memory is activated. The control device according to any one of (2) to (4), which is updated.
(6) The control according to any one of (2) to (5), wherein the evaluation information processing unit clears the evaluation information related to data for which the clone generation processing unit has performed the generation processing of the clone data. apparatus.
(7) When the nonvolatile memory is activated, the clone generation processing unit determines whether or not the refresh opportunity is close to the data stored in the nonvolatile memory based on the evaluation information, and the refresh opportunity The control device according to any one of (1) to (6), wherein the clone data generation process is performed on data determined to be close to each other.
(8) When accessing the data stored in the nonvolatile memory, the evaluation information processing unit determines whether or not a refresh opportunity is close to the data, and determines that the refresh opportunity is close When the above information is generated, the evaluation information for the data is generated,
The clone generation processing unit performs the clone data generation processing on data determined to be close to a refresh opportunity based on the evaluation information when the nonvolatile memory is activated. The control device according to any one of (6).
(9) The management information processing unit manages the data for which the clone data is generated as the main data to be accessed as the original data and the sub data as the sub data that is not to be accessed. The control device according to (1), wherein the management information is updated.
(10) When the main data has reached a refresh opportunity, the management information processing unit updates the management information so that the clone data is managed as main data to be accessed. Control device.
(11) The management information processing unit stores the management information so that one of the main data and the sub data is discarded for the data for which the clone data is generated when the nonvolatile memory is activated. The control device according to (9) or (10) to be updated.
(12) The control device according to any one of (1) to (11), wherein the evaluation information includes at least one of a bit error occurrence count, a read count, and an erase count for a predetermined data unit.

  DESCRIPTION OF SYMBOLS 1 Memory card, 2 Host apparatus, 11 Control part, 12 SRAM, 13 Device interface, 14 Buffer RAM, 15 Nonvolatile memory

Claims (14)

For data stored in a non-volatile memory and for which it is determined that refresh opportunities are close based on information for evaluation regarding refresh necessity, the same clone data as the original data is used as the original data. A clone generation processing unit for performing processing to generate in the nonvolatile memory,
A management information processing unit that updates management information so that either one of the original data or the clone data becomes an access target at the time of access execution;
An evaluation information processing unit that generates the evaluation information for the data when accessing the data stored in the nonvolatile memory;
A control device comprising:   The management information processing unit is configured to manage the clone data so that the clone data is managed as main data to be accessed and the original data is managed as sub data that is not to be accessed. The control device according to claim 1 which updates information.   3. The control according to claim 2, wherein when it is determined that the main data has reached a refresh opportunity, the management information processing unit updates the management information so that the original data is managed as main data to be accessed. apparatus.   The refresh control unit according to claim 3, further comprising: a refresh control unit configured to perform refresh operation control on the original data when it is determined that the main data has reached a refresh opportunity after the original data is the main data. Control device.   The management information processing unit is configured to update the management information so that one of the main data and the sub data is discarded with respect to data for which the clone data is generated when the nonvolatile memory is activated. Item 3. The control device according to Item 2.   The control apparatus according to claim 2, wherein the evaluation information processing unit clears the evaluation information related to data for which the clone generation processing unit has performed the generation processing of the clone data.   The clone generation processing unit determines whether or not a refresh opportunity is close based on the evaluation information for data stored in the nonvolatile memory when the nonvolatile memory is activated, and the refresh opportunity is close The control device according to claim 1, wherein the clone data generation process is performed on the data determined to be. The evaluation information processing unit determines whether or not a refresh opportunity is close to the data stored in the nonvolatile memory, and determines that the refresh opportunity is close. , Generating the evaluation information for the data,
The clone generation processing unit performs the clone data generation processing on data determined to be close to a refresh opportunity based on the evaluation information when the nonvolatile memory is activated. Control device.   The management information processing unit is configured to manage the clone data so that the original data is managed as main data to be accessed and the clone data is managed as sub data that is not to be accessed. The control device according to claim 1 which updates information.   The control device according to claim 9, wherein, when the main data reaches a refresh opportunity, the management information processing unit updates the management information so that the clone data is managed as main data to be accessed.   The management information processing unit is configured to update the management information so that one of the main data and the sub data is discarded with respect to data for which the clone data is generated when the nonvolatile memory is activated. Item 9. The control device according to Item 8.   The control device according to claim 1, wherein the evaluation information includes at least one of a bit error occurrence count, a read count, and an erase count for a predetermined data unit. Non-volatile memory;
The data stored in the non-volatile memory that is determined to be close to the refresh opportunity based on the evaluation information regarding the necessity for refresh, and the same clone as the original data is used as the original data. A clone generation processing unit for performing processing for generating data in the nonvolatile memory;
A management information processing unit that updates management information so that either one of the original data or the clone data becomes an access target at the time of access execution;
An evaluation information processing unit that generates the evaluation information for the data when accessing the data stored in the nonvolatile memory;
A storage device. For data stored in a non-volatile memory and for which it is determined that refresh opportunities are close based on information for evaluation regarding refresh necessity, the same clone data as the original data is used as the original data. A clone generation step for performing processing to generate in the nonvolatile memory,
A management information processing step for updating the management information so that either the original data or the clone data is an access target at the time of access execution;
An evaluation information processing step for generating the evaluation information for the data when accessing the data stored in the nonvolatile memory;
Control method with.

Images