JP2005242897A - Flash disk drive - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a flash disk drive whose lifetime is prevented from dropping by equalizing the number of rewriting times in order to eliminate a frequent change in address table contents and suppress an increase in the number of rewriting times in a specified area. <P>SOLUTION: A free area for data storage is prepared in advance on a flash memory 20 regardless of a logical address designated from a host machine 1, and write data from the host machine 1 is sequentially written. Although a block 21 whose data becomes invalid is erased in the background in parallel with the sequential writing, an address table 30 holding the relation between a logical address and a physical address is then stored in a nonvolatile memory 12 rewritable in a word unit, such as a RAN. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a flash disk device that is a semiconductor disk device using a flash memory that is a nonvolatile memory.

  Conventionally, as a technique related to a flash disk device, for example, there are those described in the following documents.

JP-A-2-292798 (Patent No. 326042) "Flash EEprom System" Japanese Patent Application Laid-Open No. 11-339485 “Flash EEprom System” Japanese Patent Laid-Open No. 6-124596 “Storage device using flash memory” Japanese Laid-Open Patent Publication No. 5-27924 "External Storage System Using Semiconductor Memory and Control Method Therefor"

  With the recent increase in capacity and cost of non-volatile memories, flash disk devices, which are semiconductor disk devices using non-volatile memories, have been made. A large-capacity nonvolatile memory element generally used in a flash disk device is a flash memory of an AND type / NAND type or the like.

  The flash memory is composed of a plurality of blocks, and each of these blocks is composed of a plurality of pages. A page is a unit of read (write) / write (read), and is usually 512 bytes (Bytes) in the same size as a sector which is the minimum unit of read / write in a hard disk device such as a magnetic disk device. When rewriting data in a page, normally, new data is written after erasing previous data. A block is an erasing unit (generally 16 Kbytes), and even if the rewriting is a partial page in the block, it is necessary to erase the entire block.

  In a flash disk device using such a flash memory, the data stored in the flash memory generates a bit error with a certain probability and is erased in units called blocks. There is a restriction that the lifetime of the number of erases is a short lifetime of, for example, 100,000 to 1,000,000 times.

  In order to solve these restrictions, in the techniques of Patent Document 1 and Patent Document 2, a part of the entire area of the flash memory is prepared for replacement at the time of failure, and each page has a redundant plurality of blocks. Bits are added, error correction code, bit error replacement information (perform bit replacement in a page), and block replacement information are held to solve the restriction.

  However, the conventional flash disk device has the following problems.

  When a flash disk device is used as a secondary storage device (auxiliary storage device or external storage device) of a computer system, an operating system (OS), application programs, data, etc. are stored, and arbitrary locations on the disk are read / written. The Usually, the lights in such computer systems are concentrated in a specific storage area. In this case, only blocks in the storage area in which writing is performed and blocks prepared as alternative blocks can be used for writing. For example, when a general NAND flash memory with a block rewrite limit of 100,000 times is used, if the alternative block is 20 blocks and the write is concentrated on 10 blocks, 30 blocks × 100,000 times = 3 million times This is a bottleneck in the life of the computer system.

  Further, when data is written to a certain page in a certain block in the flash disk device, it is usually necessary to erase the previous data of the corresponding page and then write new data. At this time, even if the amount of write data is small, it is necessary to erase the data of the entire block including the page (erase in units of blocks is necessary), and the data before erasure is also performed on a portion (page) where writing is not performed in the block Need to be lighted again. In the case of a flash disk device using a general NAND flash memory, reading of one page is completed in about 100 μs, but writing takes 10 ms or more.

  On the other hand, in order to solve such a problem, in Patent Documents 3 and 4, a free area is prepared in advance in the flash disk device regardless of the logical address designated by the central processing unit (hereinafter referred to as “CPU”) of the host. Then, the logical address of the write data designated by the host CPU is converted into a physical address of an empty area in the flash memory by an address table (table for converting the logical address into a real address), and the free space of this physical address is converted. A method has been proposed in which write data is sequentially written in an area. However, in this method, since the relationship between the logical address and the physical address on the flash memory changes frequently, it is necessary to change the contents of the address table each time. Therefore, in the technique of Patent Document 3, battery backup is performed and the address table is saved in a specific area of the flash memory, and in the technique of Patent Document 4, overwriting of the already written location can be performed a plurality of times. Although this is dealt with by using a special flash memory, both are disadvantageous and inconvenient.

  The present invention solves the above-described problems of the prior art, eliminates frequent changes in the contents of the address table, and suppresses an increase in the number of rewrites in a specific area, thereby reducing the lifetime by leveling the number of rewrites in the entire apparatus. It is an object of the present invention to provide a flash disk device that can prevent the above-described problem.

  In order to solve the above-described problems, the flash disk device of the present invention is configured as follows (a) to (e).

  (A) Regardless of the logical address specified by the host, an empty area for storing data is prepared in advance on the flash memory, the write data from the host is sequentially written, and the data becomes invalid in the background in parallel. The block is erased, but the address table that holds the relationship between the logical address and the physical address at that time is rewritable in a word unit such as a readable / writable memory (hereinafter referred to as “RAM”). Keep in memory. In word-unit rewriting, a unit of data that can be accessed by a memory such as a RAM in one access cycle is a word, which is usually data in units of 8 bits (16 bits) or 16 bits. Rewrite.

  (B) The address table changes are sequentially added to a specific area of the flash memory like a log. The log is the same concept as a log used in database management or the like. For example, when rewriting data in a database, at the same time as rewriting actual data, the information that “data at which location was rewritten with which value” is left. This is called a log and is an area for log. Write data can be continuously added to the disk, and even if the write data is broken, the data can be restored using the log. In the present invention, in order to store the address table, due to the characteristics of the flash memory, the lifetime decreases if it is frequently rewritten, so the changed information is added as a log instead of being erased and written. The conventional problem is solved.

  (C) When data is written to the flash memory, logical address information is held together to suppress frequent changes in the address table.

  (D) In order to level the number of rewrites so that rewrite does not concentrate only on a specific area, when data is written to a block of the flash memory, the value of the number of rewrites is simultaneously held in the block, and then erased Used to select a block to be used.

  (E) A block with a small number of rewrites when a certain difference occurs in the number of rewrites in order to effectively use a block of flash memory that holds fixed data without being updated, such as an OS Is copied to another empty block so that the block holding the fixed data can be used to level the total number of rewrites.

  According to the first aspect of the present invention, an empty table is prepared in advance in the flash memory regardless of the logical address, the write data from the host is sequentially written, and the relationship between the logical address and the physical address is maintained at that time. Is stored in a rewritable nonvolatile memory in units of words. Therefore, even when write accesses are concentrated from the host to a specific area of the flash memory, it is possible to prevent the specific area in the flash memory from being worn more than other areas. Thereby, the lifetime reduction of the whole flash disk apparatus can be suppressed.

  According to the second aspect of the present invention, since the address table is sequentially added to the specific area of the flash memory like a log, it is expensive to add the change of the address table to the flash memory. The rewritable nonvolatile memory in units of words can be deleted, and the cost can be reduced.

  According to the invention of claim 3, since only the difference of the address table is sequentially added to the specific area of the flash memory like a log, the log of the changed part of the address table (address table log) With the configuration of storing in the flash memory, even if the write from the host concentrates on a specific sector, the area on the flash memory of the address table log can be rewritten uniformly, and consumption of the specific area can be prevented.

  According to the fourth aspect of the present invention, when the data is written to the flash memory, the logical address information is held together, and the address table is configured at the start-up so that the address table is not saved. Therefore, the overhead for writing the address table and address table log to the flash memory can be eliminated, and information such as data and address is written to one page of the flash memory at the same time, so that consistency is always maintained. . Furthermore, the flash memory to be used can be selected without being restricted by the limitation on the number of times of overwriting of the flash memory.

  According to the fifth aspect of the invention, since the number of rewrites is leveled, the value of the number of rewrites is held in the block management table and used to select the block to be erased next. By managing the number of times of rewriting and using the blocks from the least number of times of rewriting, the consumption of each block can be made uniform and the life of the flash disk device can be extended.

  According to the invention of claim 6, in order to level the number of times of rewriting, when the data is written to the block of the flash memory, the value of the number of rewriting is simultaneously held in the block, and the block to be erased next is selected. Therefore, rewriting the block management table on the flash memory is limited to when a block failure occurs, and normal data writing from the host does not cause rewriting of the block management table, reducing overhead. The

  According to the invention of claim 7, since the address table is compressed as a unit of read / write by combining a plurality of sectors, it is expensive to manage a plurality of sectors as one group. The amount of rewritable nonvolatile memory in units of words can be reduced. Furthermore, the read / write processing with a large size can reduce overhead by reading / writing data of one certain size (hereinafter referred to as “chunk”) to the flash memory together.

  According to the invention of claim 8, in order to effectively use a block of flash memory that holds fixed data without being updated, when a certain difference occurs in the number of rewrites, Since a small block of data was copied to another free block so that this block could be used, if there was a flash memory area that was not updated from the host after the initial installation, the flash memory in which that data was written This area can also be used as a data area that is sequentially updated from the host during operation. Therefore, the total amount of rewriting of the flash disk device increases.

  The flash disk device of the present invention includes a flash memory and control means. The flash memory has a plurality of blocks each composed of a plurality of data storage pages with physical addresses, and data is written and read in units of pages, and data is erased in units of blocks. After that, data is written to a predetermined page in the erased block. The control means controls writing and reading of data to and from the flash memory using an address table for converting a logical address designated by a host into the physical address. The address table is stored in a rewritable nonvolatile memory in units of words or the flash memory. Alternatively, the address table is generated by the control means on the basis of logical address information written in the flash memory.

(Constitution)
1A and 1B are schematic configuration diagrams of a flash disk device showing an embodiment of the present invention. FIG. 1A is an overall configuration diagram, and FIG. 2 is a schematic configuration diagram of the flash memory in FIG.

  In FIG. 1A, the host machine 1 such as a host CPU is referred to as a host interface (hereinafter referred to as “I / F”). The flash disk device 10 used in this embodiment is connected by the bus 2. The flash disk device 10 has, for example, a microprocessor unit (hereinafter referred to as “MPU”) 11 which is a control means for controlling the entire device. The MPU 11 includes a RAM 11a, a read-only memory (hereinafter referred to as “ROM”) 11b, an input / output (hereinafter referred to as “I / O”) port 11c, and the like, to which a nonvolatile memory 12 is connected. Yes. The nonvolatile memory 12 is access-controlled by the MPU 11 and has a memory structure (FRAM, MRAM) that can be rewritten in units of words, such as a RAM that can be easily controlled to hold a management table or list. Etc.).

  The flash disk device 10 is provided with a host I / F controller 13 connected to the host I / F bus 2. The host I / F controller 13 is a circuit that processes the protocol of the host I / F and exchanges data with the host machine 1. A buffer memory 15 is connected to the host I / F controller 13 via the internal data bus 14. ing. The buffer memory 15 is a memory that temporarily stores and processes read / write data from the host machine 1. A flash memory controller 16 is connected to the internal data bus 14, and one or more flash memories 20 are connected to the flash memory controller 16 via a flash memory bus 17.

  The flash memory controller 16 is a circuit that controls access to the flash memory 20 in accordance with an instruction from the MPU 11, and includes a data reading unit 16a, a data writing unit 16b, a block erasing unit 16c, and the like. The one or more flash memories 20 are memories that store write data from the host machine 1 in a nonvolatile manner.

  In FIG. 1B, the flash memory 20 has a plurality of blocks 21-N, 21- (N + 1),. Each block 21 has a plurality of pages 22-0, 22-1,..., 22-31 for storing data, which are read / write units. One page 22 has a storage capacity of, for example, 512 bytes. In such a flash memory 20, it is necessary to erase one entire block even if rewriting is performed on some pages 22 in the block 21.

  FIG. 2 shows the first embodiment of the present invention and is a logical configuration diagram of the flash disk device 10 of FIG.

  A logical sector number (hereinafter, this number is referred to as “No.”) 0, 1, 2,..., J, which is a sector address (that is, a logical address) 25 that is visible (that is, designated) from the host machine 1. An address table 30 is stored in the nonvolatile memory 12 shown in FIG. 1 for conversion into a page address that is a physical address on the flash memory 20. The address table 30 stores a plurality of address conversion data 31-0, 31-1, 31-2,..., 31-K for converting a logical sector into a physical page. Further, in the nonvolatile memory 12, a free page list 32 for managing free pages in the flash memory 20 and blocks 21-0, 21-1,..., 21-L in the flash memory 20 are managed. A block management table 33 is stored. The plurality of blocks 21-0 to 21-L stored in the flash memory 20 have a plurality of pages 22-0, 22-1,.

  In the first embodiment, an address table 30 is prepared in the non-volatile memory 12, and a logical sector No. that is a sector address that can be seen from the host machine 1. To the physical page number on the flash memory 20. Keep a mapping to. A plurality of extra pages 22-i are prepared in the flash memory 20 and registered in the free page list 32 for management.

(Operation)
A read operation (1) from the flash memory 20, a write operation (2) to the flash memory 20, a garbage collection operation (3) of the flash memory 20, and other operations (4) will be described.

(1) Read Operation FIG. 3 is a flowchart showing the read operation of the flash disk device 10 of FIG.

  When reading data stored in the flash memory 20 in accordance with an instruction from the host machine 1, the following processing is performed under the control of the MPU 11.

  When the read process is started, in step S1, the logical sector No. which is the logical address 25 designated by the host machine 1 is displayed. (For example, logical sector No. 1) is obtained. This logical sector No. 1 is sent to the MPU 11 via the host I / F controller 13. In step S <b> 2, the MPU 11 reads the specified logical sector No. from the address table 30 in the nonvolatile memory 12. 1 corresponding to the physical page number. (For example, the address translation data 31-1) is obtained, and the page No. (For example, 22-0) is obtained. This page No. Is sent to the flash memory controller 16.

  In step S3, the data reading unit 16a in the flash memory controller 16 determines the page number. The stored data (22-0) is read out and temporarily stored in the buffer memory 15. In step S 4, the read data stored in the buffer memory 15 is transferred to the host machine 1 via the host I / F controller 13 and the host I / F bus 2. As a result, the read process ends.

(2) Write Operation FIG. 4 is a flowchart showing the write operation of the flash disk device 10 of FIG.

  When the write data from the host machine 1 is written to the flash memory 20 in accordance with the instruction from the host machine 1, the following processing is performed under the control of the MPU 11.

  When the write process is started, in step S 11, the host I / F controller 13 sends the logical sector No. corresponding to the logical address 25 from the host machine 1. (For example, logical sector No. 1) and write data are received. The received logical sector No. 1 and write data, the logical sector No. Is sent to the MPU 11 and the write data is temporarily stored in the buffer memory 15. In step S12, the MPU 11 determines the page number in the usable block (for example, 21-L) from the free page list 32 in the nonvolatile memory 12. (For example, 22-i) is obtained and sent to the data writing unit 11b in the flash memory controller 13.

  In step S13, the data writing unit 11b converts the write data stored in the buffer memory 15 into the page number. Write to the flash memory location (22-i). In step S14, the MPU 11 reads the page number from the empty page list 32. In addition to deleting (22-i), in the next step S15, the mapping on the address table 30 is changed (that is, the address conversion data is updated), and the write process is terminated.

(3) Garbage Collection Operation FIG. 5 is a flowchart showing the garbage collection operation of the flash disk device 10 of FIG.

  Since there are no free pages in the flash memory 20 when writing is performed, garbage collection is periodically performed in order to secure a certain amount of free pages. When there is no access from the host machine 1 for a certain period of time, or when the number of empty pages in the flash memory 20 becomes less than a certain amount, the following garbage collection process is performed under the control of the MPU 11.

  When the garbage collection process is started, the MPU 11 creates a list (temporary page list) indicating the state of each page in step S21. In step S22, an unused flag is set in the temporary page list corresponding to a page that is not used in the address table 30. In step 23, the block 21 in which all the pages 22-0 to 22-M are unused is searched from the temporary page list, erased by the block eraser 16c, and the flags of the corresponding pages 22-0 to 22-M are erased. And add it to the free page list 32. In step S24, the MPU 11 determines whether or not the number of empty pages 22 is equal to or larger than a predetermined amount. If yes, the garbage collection process is terminated. If no, the process proceeds to step S25 via the connector 1.

  In step S25, the MPU 11 searches the temporary page list for a block 21 with many unused pages 22. In step S26, the valid page 22 data of the block 21 is read, and an unused flag is set in the temporary page list. In step S27, the MPU 11 finds and writes out the page 22 in which the read data can be written in the empty page list 32, updates the empty page list 32 and the address table 30, and returns to step S23 via the connector 2. Repeat the above process.

  As described above, in the garbage collection process, the invalid page 22 is searched on the address table 30. If all the pages 22-0 to 22-M in the block 21 are invalid, the block 21 is searched. Erasing and registering the pages 22-0 to 22-M in the empty page list 32. If a block 21 in which all the pages 22-0 to 22-M are invalid is not found, data of some valid pages 22 is transferred to the empty pages 22, and the address table 30 and the empty page list 32 are stored. Align. When all the pages 22-0 to 22-M are invalidated in this way, the block 21 is erased by the block eraser 16c.

(4) Other Operations A failure of the block 21 of the flash memory 20 occurs at the time of erasing or writing, but is detected by returning an error status from the flash memory 20 to the MPU 11. In the MPU 11, if there is a failure in the block 21, a flag indicating that there is a failure is set in the block management table 33, and thereafter, the block 21 is managed not to be used.

  The address table 30, the free page list 32, and the block management table 33 are stored in the nonvolatile memory 12 such as FRAM and MRAM that can be rewritten in units of words. However, the data will not be lost.

(effect)
In the first embodiment, even when the write access is concentrated from the host machine 1 to a specific area in the flash memory 20, the specific area in the flash memory 20 is prevented from being worn more than other areas. I can do it. Thereby, the lifetime reduction of the whole flash disk apparatus can be suppressed.

  In the first embodiment, when the capacity of the entire flash memory is increased, the address table 30 stored in the nonvolatile memory 12 is also increased. However, a rewritable nonvolatile memory in units of words generally has a high bit cost. The flash disk device 10 becomes expensive. In a normal computer system, there are cases where the unit of access to the disk device is determined to be 4 sectors, 16 sectors, etc. In this case, if the disk management is performed on a sector basis (ie, page basis), the overhead is reduced. growing. Here, the overhead refers to, for example, a storage area and processing time that are not directly related when the computer executes a user program.

  Therefore, in the second embodiment, in order to reduce overhead, a configuration is adopted in which a plurality of sectors are collectively managed as a disk.

(Constitution)
FIG. 6 shows a second embodiment of the present invention, and is a logical configuration diagram of the flash disk device 10 of FIG.

  In order to collectively manage a plurality of sectors (or pages) in units of chunks, read / write from the host machine 1 to the flash memory 20 is performed in units of chunks (that is, a plurality of sectors in the logical address 25A of the host machine 1). The logical sectors are collectively accessed in units of logical chunks). In the nonvolatile memory 12, an address table 30A and an empty chunk list 32A are stored. The address table 30A stores a plurality of pieces of address conversion information for converting a logical chunk designated by the host machine 1 into a physical chunk that is a physical address on the flash memory 20 (that is, mapping to the chunk). have). The empty chunk list 32A is a list of physical chunks to which no data is written.

  The flash memory 20 has a plurality of flash blocks (hereinafter referred to as “Flash blocks”) 21A-0, 21A-1,. Each of the Flash blocks 21A-0, 21A-1,... Has a plurality of physical chunks in which a plurality of pages are collected.

(Operation)
A read operation (1) from the flash memory 20 and a write operation (2) to the flash memory 20 will be described.

(1) Read Operation FIG. 7 is a flowchart showing the read operation of the flash disk device 10 of FIG.

  When reading data stored in the flash memory 20 in accordance with an instruction from the host machine 1, the following processing is performed under the control of the MPU 11.

  When the read process is started, the logical sector No. designated by the host machine 1 is checked in step S31. Is obtained by MPU11. In step S32, the MPU 11 checks the logical sector number. Logical chunk No. including From the address table 30A, the physical chunk No. Get. In step S <b> 33, the MPU 11 reads the physical chunk data in the flash memory 20 and temporarily stores it in the buffer memory 15. In step S34, the logical sector number requested by the host machine 1 is set. The data corresponding to is transferred from the buffer memory 15 to the host machine 1, and the read process ends.

(2) Write Operation FIG. 8 is a flowchart showing the write operation of the flash disk device 10 of FIG.

  When the write data from the host machine 1 is written to the flash memory 20 in accordance with the instruction from the host machine 1, the following processing is performed under the control of the MPU 11.

  When the write process is started, the logical sector No. is received from the host machine 1 in step S41. And the data to be written are received by the MPU 11. In step S 43, the MPU 11 reads the physical chunk data in the flash memory 20 and temporarily stores it in the buffer memory 15. In step S44, the MPU 11 checks the logical sector number of the read data. The portion corresponding to is rewritten with the data received from the host machine 1 (that is, modification that is partial rewriting is performed). In step S45, the MPU 11 searches the empty chunk list 32A, writes back the data to the physical chunk, and updates the empty chunk list 32A and the address table 30A. As described above, in the case of writing, the physical chunk is read / modified / written on the buffer memory 15, and the writing process is completed.

(effect)
In the second embodiment, by managing a plurality of sectors as one group, it is possible to reduce the use amount of the rewritable nonvolatile memory 12 in expensive word units. Furthermore, the read / write processing with a large size can reduce overhead by reading / writing data for one chunk to the flash memory 20 together.

  In the first and second embodiments, in order to store the address tables 30 and 30A and the block management table 33, the rewritable nonvolatile memory 12 in units of expensive words is used. As a measure for reducing the cost, it is conceivable to delete the nonvolatile memory 12. However, since the address tables 30 and 30A are frequently updated, the area in which the address tables 30 and 30A are stored becomes a bottleneck for rewriting restriction, and the rewriting overhead occurs.

  The flash memory 20 has a feature that the value of the erased state is “1” and the value is held depending on whether “0” is written or left as it is. Therefore, using this feature, the third embodiment attempts to reduce the cost by updating the address table 30.

(Constitution)
FIG. 9 shows the logical configuration of the address table 30B used in the flash disk device 10 of FIG. 1 (A), showing Embodiment 3 of the present invention.

  The address table 30B is formed in the RAM 11a, which is a volatile memory in the MPU 11, and the contents are stored in the allocated area of the flash memory 20. That is, for each logical chunk N, N + 1, N + 2,. Prepare multiple words to hold First, one fixed physical chunk is held, and the others are filled with the value “1”. Such an address table 30B is stored in the allocated area of the flash memory 20 as it is.

(Operation)
FIG. 10 is a flowchart showing the update operation of the address table 30B of FIG.

  When the address table 30B is updated by rewriting, the following processing is performed under the control of the MPU 11.

  When the update process is started, in step S51, the MPU 11 determines the physical chunk No. in the logical chunk (for example, chunk No. N + 2 in FIG. 9) in the address table 30B on the RAM 11a. It is determined whether or not there is a surplus place to add. In the logical chunk N + 2, the physical chunk No. In this case, in step S52, in the address table 30B on the RAM 11a, the physical chunk number is added to the row of the logical chunk (horizontal axis in FIG. 9). (For example, physical chunk data 010...) Is added. Thereafter, in step S53, the page of the flash memory 20 corresponding to the updated location of the RAM 11a is overwritten. That is, a part of the address table 30B including the updated portion is written in the minimum unit of writing to the flash memory 20. As a result, due to the characteristics of the flash memory 20, the physical chunk No. Can be saved, and the update process ends.

  In step S51, if there is no place to hold the physical chunk in some of the logical chunks N, N + 1, N + 2,..., The process proceeds to step S54. In step S54, the MPU 11 checks the block number on the flash memory 20 corresponding to the location to be updated on the RAM 11a. (For example, 21-0 in FIG. 2) is obtained. In step S55, the MPU 11 checks the block No. The area of the address table 30B on the RAM 11a corresponding to 21-0 (for example, the logical chunk N in FIG. 9) is cleared except for the latest physical chunk ("1" is written). Thereafter, in step S56, the MPU 11 erases the block 21-0 of the flash memory 20, and writes the address table 30B on the corresponding RAM 11a.

  As described above, when there is no place to hold physical chunks in some rows, in the processing of steps S54 to S56, the initial state on the RAM 11a (a state in which only one physical chunk is held) is set. The flash memory 20 is erased / written. At this time, each item of the address table 30B corresponding to the size of the erase unit of the flash memory 20 is set to the initial state. Thereby, the update process ends.

  Here, the block management table (for example, 33 in FIG. 2) for managing the blocks 21-0,... Of the flash memory 20 is updated only when a block failure occurs. The 20 allocated areas are erased and written.

  When the flash disk device 10 is started up, the MPU 11 reads the address table 30B and the block management table (for example, 33 in FIG. 2) stored in the flash memory 20 and puts them on the RAM 11a. The free page list (for example, 32 in FIG. 2) is configured on the RAM 11a by searching the address table 30B to find a physical chunk that is not allocated.

(effect)
In the third embodiment, by adding the change of the address table 30B to the flash memory 20, the expensive rewritable nonvolatile memory 12 can be deleted and the cost can be reduced.

  In the third embodiment, when rewriting concentrates on a specific sector, the address table 30B on the flash memory 20 is frequently erased / written, which becomes a bottleneck in the number of rewriting / performance.

  Therefore, in the fourth embodiment, when the address table 30B is updated, only the changed part is saved like a log.

(Constitution)
FIG. 11 is a logical configuration diagram of an address table log used in Embodiment 4 of the present invention.

  In the fourth embodiment, the address table 30C holds the basic format (the address table 30 in FIG. 2) in the first embodiment on the RAM 11a and the flash memory 20 in the initial state. Furthermore, as shown in FIG. 11, an address table log 34 having conversion data from logical chunks to physical chunks is configured on the RAM 11a, and an area for storing this is prepared on the flash memory 20. The address table 30C and the address table log 34 are updated by the MPU 11.

(Operation)
Under the control of the MPU 11, an update operation (1) of the address table 30C and an operation (2) performed periodically are performed as follows.

(1) Update Operation of Address Table 30C FIG. 12 is a flowchart showing the update operation of the address table 30C of FIG.

  When the update process of the address table 30C is started, in step S61, a physical chunk No. is stored in the address table 30C on the RAM 11a. Update. In step S61, a pair of a logical chunk and a physical chunk is added to the “1” area in the address table log 34 on the RAM 11a. In step S63, the page on the flash memory 20 corresponding to the added log is overwritten.

  Thus, when the address table 30C is updated, the items on the RAM 11a are updated, but the items on the flash memory 20 are not updated. Instead, the updated logical chunk / physical chunk pair is added to the address table log 34 on the RAM 11 a as one item of the log, and is written to the flash memory 20. Thereby, the update process ends.

(2) Operations Performed Periodically FIG. 13 is a flowchart showing operations performed periodically in FIG.

  When the operation process to be performed periodically is started, the areas of the address table 30C and the address table log 34 on the flash memory 20 are deleted in step S65. In step S66, the data of the address table 30C on the RAM 11a is written to the flash memory 20. Thereafter, in step S67, the area of the address table log 34 on the RAM 11a is cleared ("1" is written).

  As described above, the data in the address table 30C on the RAM 11a is periodically erased / written to the flash memory 20, and at the same time, the address table log 34 on the RAM 11a and the flash memory 20 is cleared ("1" on the RAM 11a). And the flash memory 20 is erased). Thereby, the process ends.

  When the flash disk device 10 is started up, the address table 30C stored in the flash memory 20 is first read onto the RAM 11a, and then the address table log 34 on the flash memory 20 is read out on the RAM 11a. By updating the value at, the address table 30C immediately before the power is turned off can be restored.

(effect)
In the fourth embodiment, since the log (address table log 34) of the changed part of the address table 30C is stored in the flash memory 20, even when writes from the host machine 1 are concentrated on a specific sector, Since the area on the flash memory 20 of the address table log 34 is rewritten uniformly, it is possible to prevent the specific area from being consumed.

  In the configurations of the first to fourth embodiments, there is an overhead in updating the address tables 30, 30A, 30B, 30C and the address table log 34 on the flash memory 20, and the data update and the address tables 30, 30A, 30B, Since the update timings of 30C and the address table log 34 are different, there are cases where the alignment cannot be achieved when the power is turned off during that time. Further, depending on the type of flash memory 20, there is a restriction on the number of page overwrites.

  In order to solve these problems, the fifth embodiment employs a configuration in which the address tables 30, 30A, 30B, 30C on the flash memory 20 and the address table log 34 are not required.

(Constitution)
FIGS. 14A and 14B are logical configuration diagrams showing Embodiment 5 of the present invention. FIG. 14A is a configuration diagram of an address table on the RAM, and FIG. It is a block diagram which shows the content of 1 page of.

  14A, the address table 30D on the RAM 11a includes the physical chunk numbers of FIG. 6, FIG. 9, and FIG. In addition, an update No. indicating how many times the logical chunk has been updated since initialization. Save.

  In FIG. 14B, the redundancy bit of each page 22A of the flash memory 20 is normally 16 bytes in the case of a NAND flash memory, for example, and management data is stored there. The management data includes a logical chunk No. corresponding to this page, such as an error correction code of the data. , And this logical chunk No. Update No. corresponding to Save. The error correction code can also include a code for management data.

(Operation)
The operation (1) at the time of starting up the flash disk device 10 and the operation (2) at the time of data writing will be described.

(1) Operation at Startup FIG. 15 is a flowchart showing the operation at startup of the flash disk device 10 of FIG.

  When the flash disk device 10 is started up, the following processing is performed under the control of the MPU 11a.

  When the start-up process is started, in step S71, the physical chunk No. I is set to the value 0, and the last physical chunk No. Search for M. In step S72, it is determined whether I <M. If yes, the start-up process is terminated, and if no, the process proceeds to step S73. In step S73, the logical chunk No. of the first page of physical chunk I. And update No. Read out. In step S74, the ethical chunk number of the address table 30D. I is registered and the update No. If the answer is no, the program proceeds to step S75. If the answer is yes, the program proceeds to step S76. In step S75, the read logical chunk No. And update No. The address table 30D is updated at step S76. In step S76, the physical chunk No. +1 is added (incremented) to I, the process returns to step S72, and the above process is repeated.

  As described above, when the flash disk device 10 is started up, the management data of each page 20A of the flash memory 20 is read and the stored logical chunk No. In the address table 30D corresponding to the physical chunk No. And update No. Write. In the process of reading each page 20A, the logical chunk No. already created on the address table 30D. If there is the same as the update No. The larger one is adopted. When all the pages 20A have been processed, the address table 30D is completed. If the size of the chunk is larger than the size of the page 20A of the flash memory 20, the logical chunk No. in the management data of each page 20A is displayed. , And update No. By writing only to the first page 20A of the chunk and reading only the page 20A at the time of start-up, the speed can be increased.

(2) Operation at Data Writing FIG. 16 is a flowchart showing an operation at the time of data writing in the flash disk device 10 of FIG.

  At the time of data writing, the following processing is performed under the control of the MPU 11a.

  When the data writing process is started, the MPU 11a receives the logical chunk No. instructed from the host machine 1. In step S81, the physical chunk No. to be written is changed. And update No. Is obtained from the address table 30D. In step S82, the update No. In step S83, the logical chunk No. is added to the data to be written. And update No. Add management data consisting of In step S84, the data is written to the flash memory 20, the empty chunk list on the RAM 11a is updated, and the address table 30D is updated with a new update number. To update the data writing process.

  As described above, when data is written, the physical chunk No. of the address table 30D. As well as update No. Is incremented. Then, when writing data to the page 20A of the flash memory 20, management data is prepared together and transferred to the flash memory 20.

(effect)
In the fifth embodiment, the redundant chunk portion (management data portion) of the page 20A of the flash memory 20 is assigned a logical chunk No. corresponding to the page 20A. And update No. , The overhead for writing the address tables 30, 30A, 30B, 30C and the address table log 34 to the flash memory 20 as in the first to fourth embodiments can be deleted, and information such as data and addresses can be deleted. Are simultaneously written in one page 22A of the flash memory 20, so that the matching is always maintained. Furthermore, the flash memory to be used can be selected without being restricted by the limit on the number of times of overwriting of the flash memory 20.

  In the fifth embodiment, since empty blocks are used in order (rotation), there is a difference in the number of times of rewriting of individual blocks in the flash memory 20, and a failure block may occur earlier. There is a risk that performance will deteriorate.

  In the sixth embodiment, therefore, a block with a small number of rewrites is prioritized as the use order of empty blocks.

(Constitution)
17A and 17B are logical configuration diagrams of the block management table showing the sixth embodiment of the present invention. FIG. 17A is a logical configuration diagram of the block management table on the RAM, and FIG. B) is a logical configuration diagram of a block management table on the flash memory.

The block management table 33A in FIG. 17A is provided on the RAM 11a and has a flag, a rewrite number, and a rewrite number SUB of each block to be managed. A block management table 33B in FIG. 17B is provided on the flash memory 20, and has a flag and the number of rewrites of each block to be managed. The flag indicates whether the block is in failure. The number of rewrites holds the number obtained by dividing the actual number of rewrites by some number α. For example, the number α is 100. The number of rewrites SUB holds the number of rewrites from 0 to α. As a result, the actual number of rewrites of the block is
Number of rewrites x α + number of rewrites SUB
Indicated by

(Operation)
FIG. 18 is a flowchart showing an operation at the time of block erasure in the garbage collection in the flash disk device 10 of FIG.

  When a block is erased by garbage collection, the following processing is performed under the control of the MPU 11.

  When the processing is started, in step S91, the erased block rewrite number SUB in the block management table 33A on the RAM 11a is incremented. In step S92, it is determined whether or not the number of rewrites SUB> α. If no, the process ends. If yes, the process proceeds to step S93. In step S93, the rewrite number in the block management table 33A on the RAM 11a is incremented, and the rewrite number SUB is cleared. In step S94, the block management table 33B on the flash memory 20 is erased, the block management table 33A on the RAM 11a is written, and the process ends.

  Thus, in the sixth embodiment, when the block is erased by the garbage collection, the rewrite number SUB is incremented. When the block is written / erased several times and the rewrite number SUB reaches a certain number α (= 100), the rewrite number is incremented and the rewrite number SUB is cleared. At this time, the flag and the number of rewrites in the block management table 33A on the RAM 11a are erased / written to the area of the block management table 33B in the flash memory 20.

  Although the value of the number of rewrites SUB is lost when the power is turned off, there is no problem because it is in the range of 0 to α in the rewrite count specification of 100,000 for each block. When writing to data, the block is used from the block with a small number of rewrites calculated from the rewrite number and the rewrite number SUB.

(effect)
In the sixth embodiment, by managing the number of times of rewriting of each block and using the blocks from the least number of times of rewriting, the consumption of each block can be made uniform and the life of the flash disk device 10 can be extended.

  In the sixth embodiment, when the number of writes from the host machine 1 is large, the block management table 33B on the flash memory 20 is rewritten at a rate of once for writing the data of (block data size × α). , It becomes a big overhead.

  Thus, in the seventh embodiment, information related to the number of rewrites is held in the management data in the page of the flash memory 20.

(Constitution)
FIGS. 19A to 19C are block management table and page logical configuration diagrams showing Embodiment 7 of the present invention, and FIG. 19A is a logical configuration diagram of the block management table on the RAM. (B) is a logical configuration diagram of a block management table on the flash memory, and (C) is a logical configuration diagram of pages on the flash memory.

  The block management table 33C in FIG. 19A is provided on the RAM 11a and has a flag and the number of rewrites for each block to be managed. The number of rewrites represents the real number of rewrites of the block. The rewriting number SUB is not required for the management table 33A shown in FIG.

  The block management table 33D in FIG. 19B is provided on the flash memory 20 and has only flags for each block to be managed. The flag indicates a block failure and is updated only when there is a block failure. The number of rewrites is not required for the block management table 33B shown in FIG.

  In each page 22B on the flash memory 20 in FIG. 19C, the number of rewrites is added to the management data with respect to the page 22A shown in FIG. 14B of the fifth embodiment. That is, the redundant bit of each page 22B is normally 16 bytes in the case of a NAND flash memory, for example, and management data is stored there. The management data includes a logical chunk No. corresponding to this page, such as an error correction code of the data. , This logical chunk No. Update No. corresponding to , And the newly added rewrite number is saved.

(Operation)
An operation (1) at the time of data writing and an operation (2) at the time of start-up will be described in the flash disk device 10 of FIG.

(1) Operation at Data Writing FIG. 20 is a flowchart showing the operation at the time of data writing in the flash disk device 10 of FIG.

  At the time of data writing, the following processing is performed under the control of the MPU 11a.

  When the data writing process for the logical address given from the host machine 1 is started, the physical chunk No. to be written is read from the address table (for example, 30D in FIG. 14A) on the RAM 11a in step S101. And update No. And the number of rewrites is obtained from the block management table 33C on the RAM 11a. In step S102, the update No. And the rewrite number is incremented. In step S103, as shown in FIG. 19C, the logical chunk No. , Update No. And the number of rewrites. Thereafter, in step S104, the data is written into the flash memory 20, and the free chunk list (for example, 32A in FIG. 6) is updated. Further, the address table (for example, 30D in FIG. To update the number of rewrites in the block management table 33C, and the data write operation is terminated.

  As described above, when data is written to a block of the flash memory 20, the number of rewrites of the block obtained from the block management table 33C is added to the management data in the page 22B on the flash memory 20.

  When the block is erased by garbage collection, the number of rewrites of the block in the block management table 33C on the RAM 11a is incremented.

(2) Operation at Startup FIG. 21 is a flowchart showing the operation at startup in the flash disk device 10 of FIG.

  When the flash disk device 10 is started up, the following processing is performed under the control of the MPU 11a.

  When the start-up process is started, in step S111, the block management table 33D on the flash memory 20 is read, and a flag in the block management table 33C on the RAM 11a is configured. In step S112, the physical chunk No. A value 0 is set in I and the last physical chunk No. Is set to the value M. In step S113, it is determined whether I <M. If yes, the process ends. If no, the process proceeds to step S113. In step S113, the number of rewrites of the first page of physical chunk I is read. In step S114, the read value is entered in the number of rewrites in the block management table 33C on the RAM 11a. Thereafter, the physical chunk No. The value I is incremented, the process returns to step S113, and the above process is repeated.

  As described above, when the flash disk device 10 is started up, first, the block management table 33D on the flash memory 20 is read, the flag of the block management table 33C on the RAM 11a is created, and then each page on the flash memory 20 is created. The management data of 22B is read, and the rewrite number of the block in the block management table 33C on the RAM 11a is created from the rewrite number.

(effect)
In the seventh embodiment, rewriting of the block management table 33D on the flash memory 20 is limited to when a block failure occurs, and rewriting of the block management table 33D does not occur when normal data is written from the host machine 1, so overhead is reduced. Reduced.

  When the flash disk device 10 is used in a system, generally, a program is written into the flash memory 20 only at the time of the first installation, and there is an area in the flash memory 20 that is not rewritten in actual operation. Since the block 21 on the flash memory 20 corresponding to the location is not rewritten, only a part of the rewritten blocks is consumed, which becomes a bottleneck in the number of times of rewriting as the flash disk device 10.

  Therefore, in the eighth embodiment, a configuration is adopted in which blocks that are not rewritten are replaced.

(Constitution)
FIG. 22 is a logical configuration diagram showing how fixed block contents are rewritten to empty blocks in the eighth embodiment of the present invention.

  In FIG. 22, among the logical addresses 25 specified by the host machine 1, the hatched logical addresses 25-0 and 25-i are places where rewriting is performed. Among the blocks 21 (21-0, 21-1,..., 21-14,...) Of the flash memory 20, the shaded blocks 21-0, 21-1,. This is where it is done. The blank blocks 21-2, 21-3,... Without hatching are empty blocks that are not rewritten during operation.

(Operation)
FIG. 23 is a flowchart showing the block replacement operation of the flash disk device 10.

  At the time of block replacement of the flash disk device 10, the following processing is performed under the control of the MPU 11.

  A block management table (for example, 33 in FIG. 2) is created when the flash disk device 10 is started up, and a certain amount of rewriting is performed in the entire device during one power-on ( For example, the block replacement process is started every 10,000 times, and the process proceeds to step S121. In step S121, the block management table 33 is searched, and a block in which valid data is embedded has a minimum number of rewrites (for example, 21-5) and a free block has a maximum number of rewrites (for example, 21-13). ) In step S122, it is determined whether or not the difference between the maximum value and the minimum value of the number of rewrites is a certain value (for example, 10,000 times) or more. If no, the process ends. If yes, the process proceeds to step S123.

  In step S123, data is read from the obtained block with the smallest number of rewrites (for example, 21-5), and the block 21-5 is erased to be an empty block. In step S124, data is written in the determined maximum empty block (for example, 21-13), and the number of rewrites in the block management table 33 is updated. When writing normal data, one of the empty blocks 21-2,... With a small number of rewrites is used. When rewriting this fixed block, a free block with a large number of rewrites ( For example, the data is copied for 21-13). Thereafter, in step S125, the address table (for example, 30A in FIG. 6) and the empty chunk list (for example, 32A in FIG. 6) are updated, and the block replacement process is terminated.

(effect)
In the eighth embodiment, when the area of the flash memory 20 that has not been updated from the host machine 1 after the initial installation is 50%, for example, the area of the flash memory 20 in which the data has been written is also sequentially hosted during operation. Since it can be used as an area for data updated from the machine 1, the total number of rewrites of the flash disk device 10 is doubled.

1 is a configuration diagram of a flash disk device showing an embodiment of the present invention. FIG. 1 is a logical configuration diagram of a flash disk device showing Embodiment 1 of the present invention. FIG. 3 is a flowchart showing a read operation of the flash disk device of FIG. 3 is a flowchart showing a write operation of the flash disk device of FIG. 3 is a flowchart illustrating a garbage collection operation of the flash disk device of FIG. 2. It is a logical block diagram of the flash disk apparatus which shows Example 2 of this invention. 7 is a flowchart showing a read operation of the flash disk device of FIG. 6. 7 is a flowchart showing a write operation of the flash disk device of FIG. 6. It is a logical block diagram of the address table which shows Example 3 of this invention. 10 is a flowchart showing an update operation of the address table of FIG. 9. It is a logical block diagram of the address table log used in Example 4 of this invention. It is a flowchart which shows the update operation | movement of the address table used in Example 4 of this invention. It is a flowchart which shows the operation | movement regularly implemented in Example 4 of this invention. It is a logic block diagram which shows Example 5 of this invention. It is a flowchart which shows the operation | movement at the time of starting in Example 5 of this invention. It is a flowchart which shows the operation | movement at the time of the data writing in Example 5 of this invention. It is a logical block diagram of the block management table which shows Example 6 of this invention. It is a flowchart which shows the operation | movement at the time of the block erase in the garbage collection in Example 6 of this invention. FIG. 10 is a logical configuration diagram of a block management table and a page showing Embodiment 7 of the present invention. It is a flowchart which shows the operation | movement at the time of the data writing in Example 7 of this invention. It is a flowchart which shows the operation | movement at the time of starting in Example 7 of this invention. It is a logic block diagram which shows a mode that the content of the fixed block in Example 8 of this invention is rewritten in an empty block. It is a flowchart which shows the block rewriting operation | movement in Example 8 of this invention.

Explanation of symbols

1 Host machine 10 Flash disk device 11 MPU
11a RAM
12 Nonvolatile memory 16 Flash memory controller 20 Flash memory 21, 21-0, 21-1, 21A-0, 21A-1,... Block 22, 22A, 22B, 22-0, 22-1,. Page 25, 25A Logical address 30, 30A, 30B, 30C, 30D Address table 32, 32A Free page list 33, 33A, 33B, 33C, 33D Block management table 34 Address table log

Claims (8)

It has a plurality of blocks composed of a plurality of data storage pages with physical addresses, data is written and read in units of pages, and data is erased after erasing data in units of blocks. A flash memory in which data is written to a predetermined page in the block, and
A rewritable non-volatile memory in word units that holds an address table in which information for converting a logical address designated by a host into the physical address is recorded;
Control means for controlling the entire apparatus based on a command from the host,
1. A flash disk device comprising: an empty area for storing data in advance in the flash memory regardless of the logical address; and data from the host is sequentially written in the prepared empty area. It has a plurality of blocks composed of a plurality of data storage pages with physical addresses, data is written and read in units of pages, and data is erased after erasing data in units of blocks. A flash memory in which data is written to a predetermined page in the block, and
Control means for controlling the entire apparatus based on a command from the host, and generating an address table for converting a logical address designated by the host into the physical address and storing it in a specific area of the flash memory; With
Regardless of the logical address, an empty area for storing data is prepared in advance in the flash memory, data from the host is sequentially written in the prepared empty area, and the address table is changed by the control means. The flash disk device is configured to sequentially add the changed address table to the specific area of the flash memory like a log. It has a plurality of blocks composed of a plurality of data storage pages with physical addresses, data is written and read in units of pages, and data is erased after erasing data in units of blocks. A flash memory in which data is written to a predetermined page in the block, and
Control means for controlling the entire apparatus based on a command from the host, and generating an address table for converting a logical address designated by the host into the physical address and storing it in a specific area of the flash memory; With
Regardless of the logical address, an empty area for storing data is prepared in advance in the flash memory, data from the host is sequentially written in the prepared empty area, and the address table is changed by the control means. The flash disk device is configured such that the changed portion is sequentially added to the specific area of the flash memory like a log. It has a plurality of blocks composed of a plurality of data storage pages with physical addresses, data is written and read in units of pages, and data is erased after erasing data in units of blocks. A flash memory in which data is written to a predetermined page in the block, and
Control means for controlling the entire apparatus based on a command from the host,
Regardless of the logical address designated by the host, a free area for storing data is prepared in advance on the flash memory, and the data from the host and the logical address information are sequentially written in the prepared free area, At the time of starting up the apparatus, the control means generates an address table for converting the logical address into the physical address based on the logical address information written in the flash memory. A flash disk device characterized in that access control to the flash memory is used. In the flash disk device according to any one of claims 1 to 3,
2. A flash disk device comprising: a block management table for managing the block, wherein the block management table holds a value of the number of rewrites of the block, and is used for selecting a block to be erased next. The flash disk device according to claim 4, wherein
A flash disk device characterized in that when the data is written to the block of the flash memory, the value of the number of rewrites is simultaneously held in the block and used to select a block to be erased next.   7. The flash disk device according to claim 1, wherein access control is performed in a unit of a certain size in which the plurality of pages are collected. It has a plurality of blocks composed of a plurality of data storage pages with physical addresses, data is written and read in units of pages, and data is erased after erasing data in units of blocks. A flash memory in which data is written to a predetermined page in the block, and
Control means for controlling writing and reading of data to and from the flash memory using an address table in which information for converting a logical address designated by a host into the physical address is recorded;
A flash disk characterized in that, when a predetermined difference occurs in the number of times of rewriting in the plurality of blocks, data of a block with a small number of times of rewriting is copied to another empty block and the block can be used. apparatus.

Images