(19)
(11)EP 2 058 737 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
25.10.2017 Bulletin 2017/43

(21)Application number: 08253498.3

(22)Date of filing:  28.10.2008
(51)International Patent Classification (IPC): 
G06F 12/02(2006.01)

(54)

Non-volatile memory device, memory management method, and program

Nichtflüchtige Speichervorrichtung, Speicherverwaltungsverfahren und Programm

Dispositif à mémoire non volatile, procédé de gestion de mémoire et programme


(84)Designated Contracting States:
DE FR GB

(30)Priority: 06.11.2007 JP 2007288726

(43)Date of publication of application:
13.05.2009 Bulletin 2009/20

(73)Proprietor: Sony Corporation
Tokyo 108-0075 (JP)

(72)Inventors:
  • Saeki, Shusuke
    Tokyo 108-0075 (JP)
  • Iwasaki, Satoru
    Tokyo 108-0075 (JP)
  • Ichimori, Seiya
    Tokyo 108-0075 (JP)
  • Nagahama, Hiroki
    Tokyo 108-0075 (JP)
  • Sato, Kazumi
    Tokyo 108-0075 (JP)

(74)Representative: J A Kemp 
14 South Square Gray's Inn
London WC1R 5JJ
London WC1R 5JJ (GB)


(56)References cited: : 
EP-A- 1 939 750
US-A1- 2006 059 296
US-A1- 2007 005 928
WO-A-02/058074
US-A1- 2006 271 725
US-A1- 2007 038 802
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description

    BACKGROUND OF THE INVENTION


    1. Field of the Invention



    [0001] The present invention relates to a memory device having a non-volatile memory, a memory management method, and a program which can be applied to, for example, a NAND flash memory.

    2. Description of the Related Art



    [0002] A NAND flash memory allows data to be operated electrically, by "writing", "reading" and "erasing".

    [0003] For example, since rewriting of bits by "writing" is performed in only one direction from "1" to "0", to write data anew, it is necessary to perform "writing" after setting all bits to "1" once by "erasing".

    [0004] As shown in Fig. 1, in a NAND flash memory, the minimum unit of writing/reading is a page PG, and the minimum unit of "erasing" is a block BLK as a collection of a plurality of pages.

    [0005] For example, a page PG includes 512 bytes plus spare 16 bytes, thus 528 bytes or 4224 bits. A block BLK includes 32 pages of PG0 to PG31.

    [0006] Since the minimum unit of "writing" is several tens of times larger than the minimum unit of "erasing", it is necessary to devise some measure to perform rewriting of a NAND flash in an efficient manner.

    [0007] As shown in Fig. 2, each page PG of a NAND flash memory is made up of a data area 1 into which data is written, and a spare area 2 into which additional information is written.

    [0008] Generally, in a NAND flash memory, it is necessary to perform writing to pages PG within a block BLK (including the spare area) in order from the lowest-numbered page as shown in Fig. 3A. That is, in a NAND flash memory, random writing within a block BLK is prohibited.

    [0009] Also, in a NAND flash memory, unusable blocks called defect blocks exist at the time of shipment, and it is necessary to perform reading/writing from/to the NAND flash memory while avoiding these defect blocks.

    [0010] A NAND flash memory has the above-mentioned features. In this regard, it is desirable to allow existing file systems and tools to be used as they are, or only after slight modifications, with respect to the NAND flash memory as well.

    [0011] To this end, it is necessary to allow a NAND flash memory to be used without concern for "erasing" operations or defect blocks.

    [0012] Accordingly, it is necessary to install a layer for interpreting a request from the using side of the NAND flash memory, for example, a file system, and translating the request into a command for operating the NAND flash memory (logical-physical translation layer).

    [0013] Various related techniques have been proposed in this regard (see, for example, Japanese Unexamined Patent Application Publication No. 2003-36209 and Japanese Unexamined Patent Application Publication No. 2002-32256).

    [0014] The non-volatile memory described in Japanese Unexamined Patent Application Publication No. 2003-36209 includes a mechanism for returning to the normal storage state even in the event of a system down during device operation such as a power outage. Also, the unit of logical address-physical address translation (logical-physical translation) in this non-volatile memory is a block size, and the write size is a block size.

    [0015] Also, in the NAND flash memory described in Japanese Unexamined Patent Application Publication No. 2002-32256, logical-physical translation is done in page units, which allows for more efficient rewriting of data than logical-physical translation in block units.

    [0016] While the non-volatile memory described in Japanese Unexamined Patent Application Publication No. 2003-36209 mentioned above includes a mechanism for returning to the normal storage state even in the event of a system down during device operation such as a power outage, the unit of logical-physical translation is limited to the block size, and the write size is limited to the block size.

    [0017] Therefore, even when rewriting data of a size smaller than a block size, it is necessary to perform writing to the NAND flash memory on a block size basis, which disadvantageously results in a large amount of wasteful data writes.

    [0018] The NAND flash memory described in Japanese Unexamined Patent Application Publication No. 2002-32256 allows for more efficient rewriting of data than is achieved by logical-physical translation in block units.

    [0019] However, to meet the constraint placed on the writing order of pages within a block of a NAND flash memory shown in Figs. 3A and 3B, it is necessary to additionally provide a non-volatile memory or a data area of a NAND flash memory to store management information necessary for logical-physical translation, which disadvantageously adds complexity to the system configuration.

    [0020] Also, in Japanese Unexamined Patent Application Publication No. 2002-32256, no consideration is given to a mechanism for returning to the normal storage state in the event of a system down during device operation such as a power outage.

    [0021] US 2006/0059296 discloses a memory which is erased in units of pseudo-blocks which are smaller than the physical blocks. This is achieved by writing data in corresponding virtual blocks within the physical blocks, the virtual blocks and the physical blocks having a one-to-one relationship.

    [0022] US 2007/0038802 discloses a system and method for configuration and management of a flash memory, in which each physical erase unit for the memory includes a plurality of segments each comprising a plurality of frames, which in turn comprises a plurality of pages.

    [0023] US 2007/005928 is concerned with the reduction of the time to write a non-volatile memory. This is achieved by changing the mapping between logical and physical addresses to ensure that each logical write operation causes data to be stored in a previously erased location.

    [0024] WO 02/058074 discloses the update of data in less than all pages of a non-volatile memory, by programming the new data in unused pages of the same or another block.

    [0025] EP 1 939 750 (an English language equivalent of WO 2007/033614) discloses the use of an unfinished state for checking a power cut off during writing.

    [0026] US 2006/0271725 discloses a non-volatile memory device wherein before use, the non-volatile means is divided into a plurality of sequentially addressed clusters, each cluster containing a plurality of sequentially addressed logical blocks or sections. A lookup table contains only the cluster addresses.

    SUMMARY OF THE INVENTION



    [0027] It is therefore desirable to provide a memory device, a memory management method, and a program which can achieve more efficient rewriting and improved utilization, and enables a return to a normal storage state even in the event of a system down during writing or erasure at the time of a power outage or the like.

    [0028] According to first and second aspects of the present invention, there are provided a memory device and a memory management method for a memory device in accordance with the appended claims.

    [0029] According to a third aspect of the present invention, there is provided a program for causing a computer to perform all the steps of a method according to the second aspect of the invention when the program is run on a computer.

    [0030] According to an embodiment of the present invention, the unit of logical-physical translation can be selected from a size not smaller than a page size and not larger than a block size. According to an embodiment of the present invention, by introducing information indicating the writing order of blocks (sequential numbers), management information necessary for logical-physical translation can be contained within the spare area while meeting the constraint placed on the writing order of pages within a non-volatile memory.

    [0031] According to an embodiment of the present invention, it is possible to achieve more efficient rewriting and improved utilization, and to return to a normal storage state even in the event of a system down during writing or erasure at the time of a power outage or the like.

    BRIEF DESCRIPTION OF THE DRAWINGS



    [0032] 

    Fig. 1 is a diagram showing the relationship between blocks and pages of a NAND flash memory;

    Fig. 2 is a diagram showing the general configuration of a page;

    Figs. 3A and 3B are diagrams illustrating the writing order of pages within a block;

    Fig. 4 is a block diagram showing an example of the basic configuration of a memory device according to an embodiment of the present invention;

    Fig. 5 is a diagram showing an example of the basic configuration of a physical TU and a data structure related to management information stored in a spare area;

    Fig. 6 is a diagram illustrating the structure of data on a memory according to an embodiment of the present invention;

    Fig. 7 is a diagram showing the relationship among physical blocks, pages, and TUs of a NAND flash memory;

    Fig. 8 is a diagram showing a read operation of a logical TU;

    Fig. 9 is a diagram showing a write operation of a logical TU;

    Fig. 10 is a diagram showing a fold operation according to an embodiment of the present invention;

    Fig. 11 is a diagram illustrating a specific example of how sequential numbers are assigned;

    Fig. 12 is a diagram showing an example of the relationship between a logical-physical translation table on a memory and management information (a logical TU number LTUN and a sequential number (SQN)) written in the spare area of each physical TU of a NAND flash memory;

    Figs. 13A to 13C are diagrams showing how a logical-physical translation table is built by scanning the management information of the NAND flash memory in Fig. 12 in order from a physical TU0;

    Fig. 14 is a flowchart showing a data write operation;

    Fig. 15 is a flowchart showing a fold operation according to an embodiment of the present invention;

    Fig. 16 is a flowchart showing scanning of a spare area on a block by block basis, and a procedure for restoring a logical-physical translation table and a physical TU state map;

    Figs. 17A and 17B are diagrams illustrating a process of determining whether physical TUs are new or old;

    Fig. 18 is a diagram showing an example of a power supply cutoff during writing and recovery process;

    Fig. 19 is a diagram showing an example of a power supply cutoff during erasure and recovery process;

    Fig. 20 is a flowchart of a write operation of a logical TU in a case where measures against a power supply cutoff are taken;

    Fig. 21 is a flowchart showing scanning of a spare area on a block by block basis, and a procedure for restoring a logical-physical translation table and a physical TU state map in a case where measures against a power supply cutoff are taken;

    Fig. 22 is a flowchart showing a recovery process from a power supply cutoff during writing or erasure; and

    Fig. 23 is a flowchart of a relocate process.


    DESCRIPTION OF THE PREFERRED EMBODIMENTS



    [0033] Hereinbelow, an embodiment of the present invention will be described with reference to the attached drawings.

    [0034] Fig. 4 is a block diagram showing an example of the basic configuration of a memory device according to an embodiment of the present invention.

    [0035] A memory device 10 according to this embodiment has a NAND flash memory 11 as a non-volatile memory, a memory 12 configured by, for example, a volatile memory such as a DRAM or an SRAM, and a CPU 13 as a control section.

    [0036] The NAND flash memory 11, the memory 12, and the CPU 13 are connected to each other via a bus 14.

    [0037] The NAND flash memory 11 has a memory cell array in which NAND strings with a plurality of memory cells connected in series are arranged in an arrayed fashion, and allows writing, reading, and erasing to be performed electrically.

    [0038] The minimum read/write unit of the NAND flash memory 11 is a page PG, and the minimum erase unit is a block BLK as a collection of a plurality of pages.

    [0039] For example, a page PG includes 512 bytes data area plus 16 bytes spare area, thus 528 bytes or 4224 bits. A block BLK includes 32 pages of PG0 to PG31.

    [0040] For example, a 2-Gbit chip includes 2048 blocks BLK0 to BLK2047 of 128 KB. Each block includes 64 pages of 2 KB and is 128 KB = 2 KB × 64.

    [0041] The management of access to the NAND flash memory 11 is performed in accordance with control of the CPU 13 on the basis of management information, write complete data information, CRC information, or the like written into the spare area, as well as information of a logical-physical translation table (logical address-physical address translation table) or physical translation unit (TU) state map built in the memory 12.

    [0042] It should be noted here that an access to the NAND flash memory 11 includes a memory access related to various write, read, and erase operations (including control processes corresponding to address translation, garbage collection, wear leveling, and the like).

    [0043] The memory 12 includes a ROM for program storage, and a RAM as a working memory.

    [0044] In the memory 12, under the control of the CPU 13, a logical-physical translation table (logical address-physical address translation table) described later in detail, and a physical (TU) state map are built.

    [0045] The CPU 13 functions as a control section that manages access to the NAND flash memory 11, on the basis of management information, write complete data, CRC information, or the like written into the spare area of the NAND flash memory 11, as well as information of a logical-physical translation table or a physical (TU) state map built in the memory 12.

    [0046] The CPU 13 manages access to the NAND flash memory by performing logical-physical translation in translation units (TU) each being an integer fraction of a block BLK and an integer multiple of the page size.

    [0047] The CPU 13 has a function of performing an operation control in which, of the contents of a block, only physical TUs being currently used are all copied to another block with unused physical TUs, and then the source block is erased, thereby increasing unused physical TUs additionally by a number equal to the number of invalid physical TUs included in the source block.

    [0048] In the following description, this operation is referred to as a fold operation.

    [0049] Also, by writing corresponding logical TU numbers LTUN, and sequential numbers SQN indicating the writing order of blocks into the spare area of the NAND flash memory 11, at restart, the CPU 13 builds a logical-physical translation table solely from information in the spare area of the NAND flash memory, and stores the logical-physical translation table into the memory 12.

    [0050] Also, the CPU 13 writes state information indicating the completion of writing, and a CRC of management information in the spare area into the spare area of the NAND flash memory 11, thereby allowing recovery to a normal state from a power supply cutoff during writing to or erasing of the NAND flash memory 11.

    [0051] Now, a description will be given of an example of the configuration of a physical TU and the data structure of a spare area on the NAND flash memory 11.

    [0052] Fig. 5 is a diagram showing an example of the basic configuration of a physical TU and a data structure related to management information stored in a spare area.

    [0053] As shown in Fig. 5, a physical TU is made up of a data area 21 in which normal data is stored, and a spare area 23 in which at least management information 22 of the physical TU is stored.

    [0054] The management information 22 includes a corresponding logical TU address LTUN, and a sequential number SQN as writing-order information assigned to each block BLK.

    [0055] Further, the spare area 23 has a sequential number storing area 231, and a logical TU number storing area 232. In some cases, the sequential number storing area 231 and the logical TU number storing area 232 in the spare area 23 are referred to as management area.

    [0056] As will be described later, in addition to the management information 22, information for determining whether or not writing has been completed in the TU, such as write complete data or a CRC is recorded in the spare area 23.

    [0057] In this embodiment, a block assigned with the latest sequential number is referred to as the latest block.

    [0058] Next, the structure of data on the memory 12 will be described.

    [0059] Fig. 6 is a diagram illustrating the structure of data on the memory according to this embodiment. In this example, the structure of data on the memory for managing blocks, logical TUs, and physical TUs will be described.

    [0060] On the memory 12, a logical-physical translation table LPTT, a physical TU state map STMP, and a sequential number table SQNT are formed.

    [0061] The logical-physical translation table LPTT is a table used for obtaining a corresponding physical TU number (address) PTUN from a logical TU number (address).

    [0062] The physical TU state map STMP is a map used for managing the state of each physical TU, and has four kinds of state: CLEAN, INUSE, DIRTY, and DEFECT.

    [0063] CLEAN represents an unwritten state after an erase operation.

    [0064] INUSE represents a state in which valid data referenced from the logical-physical translation table LPTT is stored (which can be said to be a valid (VALID) state).

    [0065] DIRTY represents a state in which invalid data not referenced from the logical-physical translation table LPTT is stored (which can be said to be an invalid (INVALID) state).

    [0066] DEFECT represents a state in which the physical TU in question exists within a defect block.

    [0067] The sequential number table SQNT is a table for storing sequential numbers SQN assigned to individual blocks. The sequential number table SQNT is used only at the time of the process of building the logical-physical translation table LPTT.

    [0068] Hereinbelow, a specific description will be given of the more specific function of the CPU 13 as the control section according to this embodiment, and a control function based on management information or information for determining whether or not writing has been completed.

    [0069] As described above, the characteristic feature of the CPU 13 according to this embodiment resides in logical-physical translation on a TU (translation unit) basis.

    [0070] Fig. 7 is a diagram showing the relationship between physical blocks, pages, and TUs of a NAND flash memory.

    [0071] The size of a TU is an integer fraction of the block size and an integer multiple of the page size.

    [0072] The actual TU on a NAND flash memory device is referred to as physical TU, and a TU provided to the using side after undergoing address translation according to an embodiment of the present invention is referred to as logical TU.

    [0073] This allows for flexible read/write size setting, such as by matching the read/write size with the read/write unit of the file system.

    [0074] To realize logical-physical translation on a per-TU basis, the CPU 13 uses the logical-physical translation table LPTT that is a table used for obtaining a corresponding physical TU number (address) from a logical TU number (address).

    [0075] The CPU 13 basically has functions of performing reading of a logical TU, writing of a logical TU, a fold, and building of a logical-physical translation table. These functions will be described below.

    <Reading of a logical TU>



    [0076] Fig. 8 is a diagram showing a read operation of a logical TU.

    [0077] When reading a logical TU, as shown in Fig. 8, the CPU 13 finds a physical TU number corresponding to a logical TU number from the logical-physical translation table LPTT, and reads the corresponding physical TU. Through the intermediation of the logical-physical translation table LPTT, a defect block can be avoided.

    <Writing of a logical TU>



    [0078] Fig. 9 is a diagram showing a write operation of a logical TU.

    [0079] Prior to a write, a physical TU P1 corresponds to a logical TU L1.

    [0080] The CPU 13 first writes data desired to be rewritten, into a physical TU P2 that has not been written yet (ST1 in Fig. 9).

    [0081] Next, the CPU 13 makes the physical TU P2 into which data has been thus written correspond to the logical TU L1 of the logical-physical translation table LPTT (ST2 in Fig. 9).

    [0082] In this case, a state in which the latest data is stored as in the physical TU P2 is referred to as valid, and a state in which old data that is to be erased is stored as in the preceding physical TU P1 is referred to as invalid.

    <Fold>



    [0083] Fig. 10 is a diagram showing a fold operation according to this embodiment.

    [0084] Since NAND flash memory 11 can be erased only on a block BLK basis, if there is another physical TU within a block BLK like the physical TU P1 in Fig. 9, it is not possible to erase that physical TU even if the physical TU is invalid.

    [0085] Therefore, as writing is repeated, the number of invalid physical TUs increases, while the number of unwritten physical TUs to which data can be written decreases.

    [0086] Accordingly, as shown in Fig. 10, in order to reduce the number of invalid physical TUs and increase the number of unwritten physical TUs, the CPU 13 performs an operation of moving only valid physical TUs to a write target block BLK (ST11), and erasing the source block (ST12).

    [0087] In this embodiment, this operation is referred to as a fold.

    <Building of a logical-physical translation table>



    [0088] The logical-physical translation table LPTT is information on the volatile memory 12, and is lost when power is turned off. The CPU 13 writes logical TU numbers into the spare area at the time of writing data into a physical TU so that, even in the above-mentioned case, the logical-physical translation table LPTT can be built from information on the NAND flash memory 11 after power is turned on.

    [0089] At startup, the CPU 13 builds the logical-physical translation table LPTT on the memory 12 by scanning management information in the spare area of each physical TU.

    [0090] It should be noted, however, that a physical TU that has become old (invalid physical TU) upon writing of a logical TU is not immediately erased. Thus, a situation occurs in which, like the physical TU P1 and the physical TU P2 in Fig. 9, for example, a plurality of physical TUs have the same logical TU number.

    [0091] To find a valid physical TU corresponding to a logical TU number at the time of building the logical-physical translation table LPTT, it is necessary for the CPU 13 to determine the last written physical TU.

    [0092] Accordingly, in this embodiment, as shown in Fig. 11, the CPU 13 performs writing in order from the lowest-ordered block while assigning sequential numbers SQN as information indicating the writing order of blocks BLK, to individual blocks BLK.

    [0093] Within the same block BLK, the CPU 13 performs writing in order from a physical TU with the lowest physical TU number so that the writing order of pages within a block in Figs. 3A and 3B is met.

    [0094] When writing to a physical TU, the CPU 13 records a sequential number SQN into the spare area in addition to a logical TU number.

    [0095] In the example of Fig. 11, S is assigned as a sequential number SQN to a block BLK BLKm, S+1 is assigned as a sequential number SQN to the next block BLK BLKm+1, S+2 is assigned as a sequential number SQN to the next block BLK BLKm+2, and S+3 is assigned as a sequential number SQN to the next block BLK BLKm+3.

    [0096] Since the CPU 13 can uniquely find the writing order of physical TUs from sequential numbers SQN when building the logical-physical translation table LPTT, it is possible to determine the last written physical TU from among physical TUs having the same logical TU number, that is, a valid physical TU.

    [0097] Next, a specific example of how a logical-physical translation table is built will be described with reference to Fig. 12 and Figs. 13A to 13C.

    [0098] Fig. 12 is a diagram showing an example of the relationship between a logical-physical translation table on the memory and management information (a logical TU number LTUN and a sequential number (SQN)) written in the spare area of each physical TU of the NAND flash memory.

    [0099] Figs. 13A to 13C are diagrams showing how a logical-physical translation table is built by scanning the management information of the NAND flash memory in Fig. 12 in order from the physical TU0.

    [0100] In Figs. 13A to 13C, newly assigned physical TUs are pointed to by arrows (→) only with respect to items that have been updated in the logical-physical translation table LPTT. Also, the reason why the physical TU indicated by (*) in the drawing is determined as DIRTY is described.

    [0101] Fig. 13A shows a state immediately after scanning of a block BLK0.

    [0102] In this case, in the block BLK0, the physical TU indicated by (*) is determined as DIRTY because a physical TU3 having the same logical TU number 4 is newer.

    [0103] Fig. 13B shows a state immediately after scanning of a block BLK1.

    [0104] In this case, in blocks BLK0 and BLK1, the physical TU of the block BLK0 which is indicated by (*) is determined as DIRTY because a physical TU4 having the same logical TU number 1 is newer.

    [0105] Fig. 13C shows a state immediately after scanning of blocks BLK2, BLK3.

    [0106] In this case, in blocks BLK3 and BLK0, the physical TU of the block BLK3 which is indicated by (*) is determined as DIRTY because a physical TU0 having the same logical TU number 5 is newer.

    [0107] Next, various specific operations based on the above-mentioned configuration will be described while focusing on the control of the CPU 13.

    [0108] First, a read operation of a logical TU will be described.

    <Reading of a logical TU>



    [0109] The CPU 13 translates the number of a logical TU to be read, into a physical TU number by referencing the logical-physical translation table LPTT stored in the memory 12. Then, the CPU 13 identifies a physical TU from which data is to be read, and reads its data area.

    [0110] Next, a write operation of a logical TU will be described.

    <Writing of a Logical TU>



    [0111] Fig. 14 is a flowchart showing a data write operation.

    [0112] In Fig. 14, a physical TU is expressed as a PTU.

    [0113] The CPU 13 first determines whether or not a write target PTU is the last TU of a block (ST101).

    [0114] If it is determined in step ST101 that a write target PTU is the last TU of a block, the CPU 13 selects one block that has not been written yet, and sets its first TU as a write target PTU (ST102).

    [0115] Then, the latest sequential number SQN is incremented y 1 (ST103).

    [0116] Next, specified data, a logical TU number (L), and the latest sequential number SQN are written to the write target PTU (ST104).

    [0117] Next, the CPU 13 updates the logical-physical translation table LPTT, and a physical TU state map STMP (ST105).

    [0118] If it is determined in step ST101 that a write target PTU is not the last TU of a block, the processing proceeds not to step ST102 but to step ST106, and the write target PTU is advanced to the next physical TU, and the processing proceeds to step ST104.

    [0119] In this way, when writing a logical TU number, writing is performed with respect to a CLEAN, lowest-numbered physical TU continuous from the last physical TU within a block BLK having the latest sequential number SQN.

    [0120] If the last physical TU is not CLEAN, an unwritten block (block of which all physical TUs are CLEAN) is randomly selected, and the sequential number SQN is advanced by 1 step and written into that block (this physical TU set as a write target is referred to as write target PTU, and a block including the write target PTU is referred to as target block).

    [0121] In step ST105, the logical-physical translation table LPTT and the physical TU state map are updated as follows.

    [0122] In the physical TU state map STMP, the state of a physical TU previously corresponding to a logical TU L in the logical-physical translation table LPTT is set to DIRTY, and the state of the write target PTU is set to INUSE.

    [0123] In the logical-physical translation table LPTT, the write target PTU is made to correspond to the logical TU L.

    [0124] Next, a fold operation will be described.

    <Fold>



    [0125] Fig. 15 is a flowchart showing a fold operation according to this embodiment.

    [0126] In the flowchart in Fig. 15, a physical TU on which fold processing is being performed is expressed as P.

    [0127] The CPU 13 selects a block including at least one DIRTY physical TU as a fold target block (ST111).

    [0128] The following processing is performed with respect to every INUSE TU within the fold target block.

    [0129] Data and management information within a physical TU are read (ST112).

    [0130] Next, by specifying data and logical TU number read in step ST112, the read contents are written. In other words, writing of a logical TU is performed (ST113).

    [0131] After the above-mentioned processing is performed with respect to every INUSE TU within the fold target block, the fold target block is erased.

    [0132] Next, building of a table at startup will be described.

    <Building of a table at startup>



    [0133] The logical-physical translation table LPTT and the physical TU state map STMP are data on the memory 12, and are lost when power is turned off.

    [0134] In this embodiment, the logical-physical translation table LPTT and the physical TU state map STMP can be restored even in that case by scanning the management information (logical TU numbers LTUN and sequential numbers SQN) written into the spare area 23 of the NAND flash.

    [0135] Fig. 16 is a flowchart showing scanning of a spare area on a block by block basis, and a procedure for restoring a logical-physical translation table and a physical TU state map.

    [0136] In the flowchart in Fig. 16, the physical TU being currently scanned is expressed as P1.

    [0137] In the scanning of a spare area of the NAND flash, the following processing is performed with respect to every block within a scan target block other than defect blocks.

    [0138] The CPU 13 reads the storing area (herein referred to as management area) 231, 232 of the management information 22 in the spare area 23 of the physical TU P1 (ST121).

    [0139] Next, it is determined whether or not the management area is unwritten (all bytes set to 0xFF) (ST122).

    [0140] If it is determined in step ST122 that the management area is unwritten, the state of the physical TU P1 is set to CLEAN (ST123).

    [0141] If it is determined in step ST122 that the management area is not unwritten, the logical TU number of the management area of the physical TU P1 is set to L. The sequential number SQN in the management area is stored into the sequential number table SQNT (ST124).

    [0142] Next, it is determined whether or not another physical TU P2 pointing to the logical TU number L is registered in the logical-physical translation table LPTT that has been built up to now (ST125).

    [0143] If it is determined in step ST125 that no another physical TU P2 is registered, in the logical-physical translation table LPTT, the physical TU number P1 is made to correspond to the logical TU number L (ST126).

    [0144] Then, the state of the physical TU P1 is set to INUSE (ST127).

    [0145] If it is determined in step ST125 that another physical TU P2 is registered, it is determined whether or not the physical TU P1 is newer than the physical TU P2 (ST128). This is determined from the sequential number SQN and the position within the block.

    [0146] If it is determined in step ST128 that the physical TU P1 is newer than the physical TU P2, in the logical-physical translation table LPTT, the physical TU number P1 is made to correspond to the logical TU number L (ST129).

    [0147] Then, the state of the physical TU P1 is set to INUSE, and the state of the physical TU P2 is set to DIRTY (ST130).

    [0148] If it is determined in step ST128 that the physical TU P1 is not newer than the physical TU P2, the state of the physical TU P1 is set to DIRTY (ST131).

    [0149] The above-mentioned processing is performed with respect to every physical TU within the scan target block.

    [0150] In this way, when building a table at startup, if there are a plurality of physical TUs that have the same logical TU number, the most recently written physical TU is determined as INUSE, and the other physical TUs are determined as DIRTY.

    [0151] The writing order can be determined from the sequential number and the position within the block.

    [0152] Figs. 17A and 17B are diagrams illustrating a process of determining whether physical TUs are new or old.

    [0153] As shown in Fig. 17A, if physical TUs are present in different blocks, which physical TU is newer or older is determined by a size comparison of sequential numbers. In this example, since the sequential number SQN S1 < S2, it is determined that the physical TU P2 is newer than the physical TU P1.

    [0154] As shown in Fig. 17B, if physical TUs are present within the same block, since the physical TU number PTUN P2 < P1, it is determined that the physical TU P1 is newer than the he physical TU P2.

    <Measures against power supply cutoff during writing to or erasing of the NAND flash memory>



    [0155] Next, a description will be given of a method devised to bring the storage state of data in the NAND flash memory 11 to the normal state following a system return process carried out after occurrence of a system down due to a power outage or an operation failure.

    [0156] When a power supply cutoff occurs during writing to the NAND flash memory 11, the value of data or management information in the physical TU being currently written to may have been left in a somewhat incomplete state. Also, when a power supply cutoff occurs during erasure of the NAND flash memory 11, data or management information in the physical TU within the block being currently erased may remain without being completely erased.

    [0157] It is thus necessary to provide a mechanism for preventing a situation where such a physical TU in a somewhat incomplete state is regarded as a valid physical TU and used, or is regarded as unwritten so that data is written over such a physical TU.

    [0158] Accordingly, the following mechanism is added.

    <Measures against power supply cutoff during writing>



    [0159] Fig. 18 is a diagram showing an example of a power supply cutoff during writing and recovery process.

    [0160] As shown in Fig. 18, immediately after writing of data and management information, state (write complete state) information 24 indicating completion of writing is additionally written into the spare area 23.

    [0161] At restart, by checking the write complete state with respect to the last written block (block having the latest sequential number), a physical TU being currently written to is detected.

    [0162] For example, as shown in Fig. 18, at restart, data and management information have been written but write complete state information has not been written with respect to the physical TU P3. Thus, the CPU 13 detects a power supply cutoff during writing.

    [0163] Then, the contents of only valid physical TUs of the block BLK1 are moved to the block BLK2 that has not been written yet.

    [0164] Then, the block BLK1 is erased.

    [0165] That is, upon detecting a physical TU that is being currently written to, the recovery process as shown in Fig. 18 is performed to erase the physical TU that is being currently written to.

    <Measures against power supply cutoff during erasure>



    [0166] Fig. 19 is a diagram showing an example of a power supply cutoff during erasure and recovery process.

    [0167] As shown in Fig. 19, a CRC 25 with respect to the management information 22 is added to the spare area 23.

    [0168] At restart, by checking the CRC of management information, it is checked whether or not the value of the management information is correct.

    [0169] Fig. 19 illustrates that even when a power supply cutoff occurs during erasure and thus the data area 21 or the spare area 23 is left in an incorrect state, the block that was being erased is erased again by a recovery process.

    [0170] For example, as shown in Fig. 19, when building a logical-physical translation table, a physical TU within this block is determined as DIRTY or CLEAN as follows, in accordance with its erase state.
    1. (a): If management information and CRC fully remain, this physical TU is DIRTY from the beginning, and is correctly determined as DIRTY.
    2. (b): If management information and CRC partially remain, the CRC check is not correct, so this physical TU is determined as DIRTY.
    3. (c): If management information and CRC have all been erased, this physical TU is determined as CLEAN.


    [0171] Then, since the block BLK has no INUSE physical TU, the block BLK is erased. In the recovery process, a block not including any INUSE physical TU is erased.

    [0172] Next, a description will be given of a difference from the operation described in the "description of operation" of the previous section, with respect to a case where measures against a power supply cutoff are taken.

    <Writing of a logical TU>



    [0173] Fig. 20 is a flowchart of a write operation to a logical TU in a case where measures against a power supply cutoff are taken.

    [0174] Comparing the process in Fig. 20 and the process in Fig. 14, the difference resides in adding a CRC to the management information to be written into the spare area, and providing, before the processing of step ST105, step ST107 in which immediately after writing of data and management information, a state indicating the completion of writing is additional written.

    <Building of a logical-physical translation table>



    [0175] Fig. 21 is a flowchart of a write operation of a logical TU in a case where measures against a power supply cutoff are taken.

    [0176] Comparing the process in Fig. 21 and the process in Fig. 16, the difference resides in proving step ST132 in which, if the result of the determination in step ST122 is negative, it is determined whether or not write complete state information is correctly written in the spare area, and step ST133 in which, if it is determined that write complete state information is correctly written, it is determined whether or not the CRC of the spare area is correct.

    [0177] After a logical-physical translation table is built, a recovery process is further performed.

    [0178] Fig. 22 is a flowchart showing a recovery process from a power supply cutoff during writing or erasure.

    [0179] For example, all blocks not including any INUSE physical TU, other than defect blocks, are erased (ST141). At this time, a recovery process for a power supply cutoff during erasure is performed.

    [0180] Next, the newest block is set as B1 and a physical TU in the newest block B1 which is not CLEAN and has the largest physical TU number is set as P1 (ST142).

    [0181] It is determined whether or not the physical TU P1 is DIRTY (ST143).

    [0182] If it is determined in step ST143 that the physical TU P1 is not DIRTY, it is determined whether or not the physical TU P1 is the last physical TU of the block B1 (ST144).

    [0183] If it is determined in step ST144 that the physical TU P1 is not the last physical TU, a physical TU that is larger in physical TU number by 1 than the physical TU P1 is set as P2. Then, the entire data area 21 and spare area 23 of the physical TU P2 are read (ST145).

    [0184] Next, it is determined whether or not all bytes of the data thus read are in the "FFh" pattern (ST146). If it is determined in step ST146 that all bytes of the read data are in the "FFh" pattern, if it has been determined in step ST144 that the physical TU P1 is the last physical TU, the processing is terminated.

    [0185] On the other hand, if it is determined in step ST143 that the physical TU P1 is DIRTY, or if it is determined in step ST146 that not all bytes of the read data are in the "FFh" pattern, the recovery process (relocate) of the block B1 is performed (ST147).

    [0186] In step ST147, a recovery process (relocate) for a power supply cutoff during writing is performed.

    [0187] Fig. 23 is a flowchart of a relocate process.

    [0188] In the flowchart of Fig. 23, the physical TU on which a relocate process is being performed is expressed by P.

    [0189] The write target PTU is set to the last physical TU of the write target block (ST1471).

    [0190] The following processing is performed with respect to every INUSE TU within a block to be relocated.

    [0191] Data and the management area (storage area where management information is stored) within a physical TU are read (ST1472).

    [0192] The read contents are written (ST1473).

    [0193] The above-mentioned processing is performed with respect to every INUSE TU within the block to be relocated.

    [0194] Then, the block to be relocated is erased (ST1474).

    [0195] It should be noted that the reason why the write target PTU is set as the last physical TU of a block in advance in step ST1471 is to set the first physical TU of an unwritten block as the physical TU from which writing is started in step ST1473.

    [0196] As described in the foregoing, according to this embodiment, the CPU 13 manages access to the NAND flash memory 11 by performing logical-physical translation in translation units (TUs) each being an integer fraction of a block size and an integer multiple of a page size. The CPU 13 has the function of performing an operation control of copying, from among the contents of a source block, only all physical blocks being currently in use to another block having unused physical TUs, and erasing the source block, thereby increasing unused physical TUs additionally a number equal to the number of invalid physical TUs included in the source block. Also, the CPU 13 has the function of writing a corresponding logical TU number, and a sequential number indicating the writing order of blocks into a spare area of the NAND flash memory, thereby building a logical-physical translation table solely from information in the spare area of the NAND flash memory at restart, and storing the logical-physical translation table into the memory 12. Therefore, the following advantages can be attained.

    [0197] The unit of logical-physical translation can be selected as a size not smaller than the page size but not larger than the block size. It is thus possible to realize logical-physical translation adapted to a unit as requested by the file system or the like.

    [0198] Since all pieces of information necessary for building management information such as a logical-physical translation table are stored in the spare area of the NAND flash, the utilization of the NAND flash can be enhanced (it is not necessary to use the data area of the NAND flash or to use an additional non-volatile memory to store information necessary for building management information such as a logical-physical translation table).

    [0199] When writing data into the NAND flash, information necessary for building management information such as a logical-physical translation table is also written simultaneously. Thus, it is not necessary to perform additional writing of management information related to logical-physical translation, thus allowing for efficient rewriting of data.

    [0200] Also, the CPU 13 has the function of writing state information indicating the completion of writing and the CRC of management information in the spare area, into the spare area of the NAND flash memory 11, thus allowing recovery to the normal state from a power supply cutoff during writing to or erasing of the NAND flash memory 11. It is thus possible to provide a mechanism for returning to the normal storage state even in the event of a system down during writing to or erasing of the NAND flash due to a power outage or the like.

    [0201] Also, the method described above in detail can be also configured such that the method is implemented as a program according to the above-mentioned procedure, and is executed by a computer such as a CPU.

    [0202] Also, such a program can be also configured such that the program is executed by being accessed from a computer in which a recording medium such as a semiconductor memory, a magnetic disc, an optical disc, or a floppy (R) disc is loaded.

    [0203] It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and alterations may occur depending on design requirements and other factors insofar as they are within the scope of the appended claims or the equivalents thereof.


    Claims

    1. A memory device (10) comprising:

    a non-volatile memory (11) which allows data to be written, read, and erased electrically and in which writing and reading are done in units of a page (PAGE) and erasing is done in units of a block (BLK) including a plurality of pages (PAGE); and

    a control section (13) that is configured to manage access to the non-volatile memory (11), wherein

    the control section (13) is configured to perform management of access to the non-volatile memory (11) by performing logical address-physical address translation (logical-physical translation) in translation units (TU) each being an integer fraction of a size of the block (BLK) and an integer multiple of a page size,

    the memory device comprising a further memory (12), wherein:

    a physical translation unit, TU, in the non-volatile memory (11) includes a data area (21) and a spare area (23); and

    the control section (13) is configured:

    to write management information including a corresponding logical translation unit number (LTUN), and a sequential number (SQN) indicating the write order of each block (BLKm, fBLKm+1, BLKm+2, BLKm+3), into the spare area (23) of the non-volatile memory (11),

    to build a logical-physical translation table (LPTT) at startup by scanning the management information in the spare area (23) of the non-volatile memory (11), and to store the logical-physical translation table (LPTT) into the further memory (12), and

    to perform a logical-physical translation process on the basis of the logical-physical translation table of the further memory (12);

    the memory device being characterised in that:

    the control section (13) is configured to restore, at the startup, a physical translation unit state map (STMP) for managing the physical translation unit (TU), in parallel with building the logical-physical translation table (LPTT) by scanning the management information in the spare area (23) of the non-volatile memory (11); and

    the physical translation unit state map (STMP) indicates whether the physical translation unit (TU) is in an unwritten, CLEAN state after an erase operation, an INUSE state in which valid data referenced from the logical-physical translation table is stored, or an INVALID state in which invalid data not referenced from the logical-physical translation table is stored, wherein:

    during the restoring of the logical-physical translation table (LPTT) and of the physical translation unit state map (STMP), if a plurality of physical translation units (TU) having the same logical TU number exist upon scanning the management information in the spare area of the non-volatile memory, the control section determines, from the sequential number (SQN) and the position within the block, the most recently written physical TU as INUSE, and determines other physical translation units (TU) as INVALID.


     
    2. The memory device according to Claim 1, wherein:

    the control section (13) is configured to perform an operation of copying, of contents of a source block, only all physical translation units (TU) that are being currently used to a block with unused physical translation units (TU), and erasing the source block to increase unused physical translation units (TU) additionally by a number equal to the number of invalid physical translation units (TU) included in the source block.


     
    3. The memory device according to Claim 2, wherein:

    when writing a logical translation unit (TU) number into the spare area, the control section performs writing to a physical translation unit (TU) within a block having the latest sequential number (SQN), said physical translation unit being a lowest-numbered, CLEAN physical TU sequential from its last physical translation unit (TU).


     
    4. The memory device according to Claim 3, wherein:

    if the last physical translation unit (TU) is not CLEAN, the control section randomly selects a block that has not been written yet, advances the sequential number by one step, and writes the sequential number into the block.


     
    5. The memory device according to Claim 4, wherein:

    the control section (13) is configured to update the physical TU state map (STMP) by setting a state of a physical TU previously corresponding to a logical TU to INVALID, and setting a state of a write target physical TU to INUSE, and to update the logical-physical translation table (LPTT) by rewriting to the write target physical TU.


     
    6. The memory device according to Claim 1, wherein
    the control section (11) is configured to determine whether physical translation (TUs) are new or old by a size comparison of sequential numbers (SQN) if the physical TUs are present in different blocks, and on the basis of physical TU numbers if the physical TUs are present within the same block.
     
    7. The memory device according to Claim 1, wherein:

    when performing a fold operation that is an operation of copying, of contents of a source block, only all physical translation units (TUs) that are being currently used to a block with unused physical translation units (TUs), and erasing the source block to increase unused physical translation units (TUs) additionally by a number equal to the number of invalid physical translation units (TUs) included in the source block, the control section reads, with respect to every INUSE TU within a fold target block, data and management information within a physical translation unit (TU), and performs writing of a logical translation unit (TU) by specifying the read data and a logical translation unit (TU) number, and then erases the fold target block.


     
    8. The memory device according to Claim 1, wherein:

    the control section (13) is configured to

    write write complete state information, which enables determination as to whether or not writing is complete, into the spare area in addition to the management information, and

    determine whether or not a power supply cutoff has occurred during writing, by checking the write complete state information at startup with respect to the last written block having the latest sequential number.


     
    9. The memory device according to Claim 1, wherein: the control section (13) is configured to
    write information indicating whether or not the management information is correct into the spare area in addition to the management information, and
    check whether or not a value of the management information is correct by checking, at startup, the information indicating whether or not the management information is correct.
     
    10. A memory management method for a memory device (10) having a non-volatile memory (11) which allows data to be written, read, and erased electrically and in which writing and reading are done in units of a page (PAGE) and erasing is done in units of a block (BLK) including a plurality of pages (PAGEs), comprising:

    performing logical address-physical address translation (logical-physical translation) in translation units (TUs) each being an integer fraction of a size of the block and an integer multiple of a page size; and

    managing access to the non-volatile memory (11) in accordance with a result of the logical-physical translation,

    wherein the memory device further comprises a further memory (12), and wherein:

    a physical translation unit, TU, in the non-volatile memory (11) includes a data area (21) and a spare area (23); and

    the method comprises the steps of:

    writing management information including a corresponding logical TU number, and a sequential number (SQN) indicating the write order of each block, into the spare area of the non-volatile memory (11),

    building a logical-physical translation table (LPTT) at startup by scanning the management information in the spare area (23) of the non-volatile memory (11), and storing the logical-physical translation table (LPTT) into the further memory (12), and

    performing a logical-physical translation process on the basis of the logical-physical translation table of the further memory (12),

    the memory management method being characterized by the steps of:

    restoring, at the startup, a physical translation unit state map (STMP) for managing the physical translation unit (TU), in parallel with building the logical-physical translation table (LPTT) by scanning the management information in the spare area (23) of the non-volatile memory (11); wherein

    the physical translation unit state map (STMP) indicates whether the physical translation unit (TU) is in an unwritten, CLEAN state after an erase operation, an INUSE state in which valid data referenced from the logical-physical translation table is stored, or an INVALID state in which invalid data not referenced from the logical-physical translation table is stored, and during the restoring of the logical-physical translation table (LPTT) and of the physical translation unit state map (STMP), if a plurality of physical translation units (TU) having the same logical TU number exist upon scanning the management information in the spare area of the non-volatile memory, determining, from the sequential number (SQN) and the position within the block, the most recent written physical TU as INUSE, and determining other physical Tus as INVALID.


     
    11. A program for causing a computer to perform all the steps of a memory management method according to claim 10 when the program is run on the computer.
     


    Ansprüche

    1. Speichervorrichtung (10), umfassend:

    einen nichtflüchtigen Speicher (11), der gestattet, dass Daten elektrisch geschrieben, gelesen und gelöscht werden, und in dem Schreiben und Lesen in Einheiten einer Seite (PAGE) erfolgen und Löschen in Einheiten eines Blocks (BLK), der eine Vielzahl von Seiten (PAGE) enthält, erfolgt; und

    einen Steuerabschnitt (13), der konfiguriert ist, Zugriff auf den nichtflüchtigen Speicher (11) zu verwalten, wobei

    der Steuerabschnitt (13) konfiguriert ist, Verwaltung des Zugriffs auf den nichtflüchtigen Speicher (11) durch Durchführen von Übersetzung logischer Adressen in physikalische Adressen (logisch-physikalische Übersetzung) in Übersetzungseinheiten (TU), die jeweils ein ganzzahliger Anteil einer Größe des Blocks (BLK) und ein ganzzahliges Vielfaches einer Seitengröße sind, durchzuführen,

    wobei die Speichervorrichtung einen weiteren Speicher (12) umfasst, wobei:

    eine physikalische Übersetzungseinheit, TU, in dem nichtflüchtigen Speicher (11) einen Datenbereich (21) und einen Reservebereich (23) enthält; und

    der Steuerabschnitt (13) konfiguriert ist:

    zum Schreiben von Verwaltungsinformationen, die eine korrespondierende logische Übersetzungseinheit-Nummer (LTUN) und eine laufende Nummer (SQN), die die Schreibreihenfolgen jedes Blocks (BLKm, BLKm+1, BLKm+2, BLKm+3) angibt, enthalten, in den Reservebereich (23) des nichtflüchtigen Speichers (11),

    zum Erstellen einer logisch-physikalischen Übersetzungstabelle (LPTT) bei der Inbetriebnahme durch Abtasten der Verwaltungsinformationen in dem Reservebereich (23) des nichtflüchtigen Speichers (11) und zum Speichern der logisch-physikalischen Übersetzungstabelle (LPTT) in den weiteren Speicher (12) und

    zum Durchführen eines logisch-physikalischen Übersetzungsprozesses auf der Grundlage der logisch-physikalischen Übersetzungstabelle des weiteren Speichers (12);

    wobei die Speichervorrichtung dadurch gekennzeichnet ist, dass:

    der Steuerabschnitt (13) konfiguriert ist zum Wiederherstellen, bei der Inbetriebnahme, einer Zustandskarte physikalischer Übersetzungseinheiten (STMP) zum Verwalten der physikalischen Übersetzungseinheit (TU) parallel mit dem Erstellen der logisch-physikalischen Übersetzungstabelle (LPTT) durch Abtasten der Verwaltungsinformationen in dem Reservebereich (23) des nichtflüchtigen Speichers (11); und

    wobei die Zustandskarte physikalischer Übersetzungseinheiten (STMP) angibt, ob die physikalische Übersetzungseinheit (TU) in einem nicht geschriebenen sauberen Zustand CLEAN nach einer Löschoperation, einem In-Gebrauch-Zustand INUSE, in dem gültige Daten, referenziert von der logisch-physikalischen Übersetzungstabelle, gespeichert sind, oder einem ungültigen Zustand INVALID, in dem ungültige, nicht von der logisch-physikalischen Übersetzungstabelle referenzierte Daten gespeichert sind, ist, wobei:

    während des Wiederherstellens der logisch-physikalischen Übersetzungstabelle (LPTT) und der Zustandskarte physikalischer Übersetzungseinheiten (STMP), wenn eine Vielzahl physikalischer Übersetzungseinheiten (TU) mit der gleichen logischen TU-Nummer nach Abtasten der Verwaltungsinformationen in dem Reservebereich des nichtflüchtigen Speichers existiert, der Steuerabschnitt aus der laufenden Nummer (SQN) und der Position innerhalb des Blocks die zuletzt geschriebene physikalische TU als INUSE bestimmt und andere physikalische Übersetzungseinheiten (TU) als INVALID bestimmt.


     
    2. Speichervorrichtung nach Anspruch 1, wobei:

    der Steuerabschnitt (13) konfiguriert ist zum Durchführen einer Operation des Kopierens von Inhalt eines Quellblocks nur sämtlicher physikalischer Übersetzungseinheiten (TU), die gegenwärtig verwendet werden, in einen Block mit nicht verwendeten physikalischen Übersetzungseinheiten (TU) und des Löschens des Quellblocks, um nicht verwendete physikalische Übersetzungseinheiten (TU) zusätzlich um eine Anzahl gleich der Anzahl ungültiger physikalischer Übersetzungseinheiten (TU), die in dem Quellblock enthalten sind, zu erhöhen.


     
    3. Speichervorrichtung nach Anspruch 2, wobei:

    beim Schreiben einer Nummer einer logischen Übersetzungseinheit (TU) in den Reservebereich der Steuerabschnitt Schreiben in eine physikalische Übersetzungseinheit (TU) innerhalb eines Blocks mit der letzten laufenden Nummer (SQN) durchführt, wobei die physikalische Übersetzungseinheit eine am niedrigsten nummerierte CLEAN physikalische TU sequenziell von ihrer letzten physikalischen Übersetzungseinheit (TU) ist.


     
    4. Speichervorrichtung nach Anspruch 3, wobei:

    wenn die letzte physikalische Übersetzungseinheit (TU) nicht CLEAN ist, wählt der Steuerabschnitt zufällig einen Block aus, in den noch nicht geschrieben wurde, schreibt die laufende Nummer um einen Schritt fort und schreibt die laufende Nummer in den Block.


     
    5. Speichervorrichtung nach Anspruch 4, wobei:

    der Steuerabschnitt (13) konfiguriert ist zum Aktualisieren der Zustandskarte physikalischer TU (STMP) durch Einstellen eines Zustands einer physikalischen TU, die vorher mit einer logischen TU korrespondierte, auf INVALID und Einstellen eines Zustands eines physikalischen Schreibziel-TU auf INUSE und zum Aktualisieren der logisch-physikalischen Übersetzungstabelle (LPTT) durch Neuschreiben in die physikalische Schreibziel-TU.


     
    6. Speichervorrichtung nach Anspruch 1, wobei
    der Steuerabschnitt (11) konfiguriert ist zum Bestimmen, ob physikalische Übersetzungseinheiten (TUs) neu oder alt sind, durch einen Größenvergleich laufender Nummern (SQN), wenn die physikalischen TUs in verschiedenen Blöcken vorhanden sind, und auf der Basis physikalischer TU-Nummern, wenn die physikalischen TUs innerhalb des gleichen Blocks vorhanden sind.
     
    7. Speichervorrichtung nach Anspruch 1, wobei:

    beim Durchführen einer Faltungsoperation, die eine Operation des Kopierens von Inhalt eines Quellblocks nur sämtlicher physikalischer Übersetzungseinheiten (TUs), die gegenwärtig verwendet werden, in einen Block mit nicht verwendeten physikalischen Übersetzungseinheiten (TUs) und des Löschens des Quellblocks, um nicht verwendete physikalische Übersetzungseinheiten (TUs) zusätzlich um eine Anzahl gleich der Anzahl ungültiger physikalischer Übersetzungseinheiten (TUs), die in dem Quellblock enthalten sind, zu erhöhen, ist, der Steuerabschnitt in Bezug auf jede INUSE TU innerhalb eines gefalteten Zielblocks Daten und Verwaltungsinformationen in einer physikalischen Übersetzungseinheit (TU) liest und Schreiben einer logischen Übersetzungseinheit (TU) durch Spezifizieren der gelesenen Daten und einer Nummer einer logischen Übersetzungseinheit (TU) durchführt und dann den gefalteten Zielblock löscht.


     
    8. Speichervorrichtung nach Anspruch 1, wobei:

    der Steuerabschnitt (13) konfiguriert ist zum

    Schreiben von Schreiben-Abgeschlossen-Zustandsinformationen, die ermöglichen, zu bestimmen, ob Schreiben abgeschlossen ist oder nicht, in den Reservebereich zusätzlich zu den Verwaltungsinformationen und

    Bestimmen, ob eine Unterbrechung der Leistungsversorgung während des Schreibens vorgekommen ist oder nicht, durch Prüfen der Schreiben-Abgeschlossen-Zustandsinformationen bei der Inbetriebnahme in Bezug auf den letzten geschriebenen Block mit der letzten laufenden Nummer.


     
    9. Speichervorrichtung nach Anspruch 1, wobei:

    der Steuerabschnitt (13) konfiguriert ist zum

    Schreiben von Informationen, die angeben, ob die Verwaltungsinformationen richtig sind oder nicht, in den Reservebereich zusätzlich zu den Verwaltungsinformationen und

    Prüfen, ob ein Wert der Verwaltungsinformationen richtig ist oder nicht, durch Prüfen, bei der Inbetriebnahme, der Informationen, die angeben, ob die Verwaltungsinformationen richtig sind oder nicht.


     
    10. Speicherverwaltungsverfahren für eine Speichervorrichtung (10) mit einem nichtflüchtigen Speicher (11), der gestattet, dass Daten elektrisch geschrieben, gelesen und gelöscht werden, und in dem Schreiben und Lesen in Einheiten einer Seite (PAGE) erfolgen und Löschen in Einheiten eines Blocks (BLK), der eine Vielzahl von Seiten (PAGEs) enthält, erfolgt, umfassend:

    Durchführen von Übersetzung logischer Adressen in physikalische Adressen (logisch-physikalische Übersetzung) in Übersetzungseinheiten (TUs), die jeweils ein ganzzahliger Anteil einer Größe des Blocks und ein ganzzahliges Vielfaches einer Seitengröße sind; und

    Verwalten von Zugriff auf den nichtflüchtigen Speicher (11) gemäß einem Ergebnis der logisch-physikalischen Übersetzung;

    wobei die Speichervorrichtung ferner einen weiteren Speicher (12) umfasst und wobei:

    eine physikalische Übersetzungseinheit, TU, in dem nichtflüchtigen Speicher (11) einen Datenbereich (21) und einen Reservebereich (23) enthält; und

    das Verfahren die folgenden Schritte umfasst:

    Schreiben von Verwaltungsinformationen, die eine korrespondierende logische TU-Nummer und eine laufende Nummer (SQN), die die Schreibreihenfolge jedes Blocks angibt, enthalten, in den Reservebereich des nichtflüchtigen Speichers (11),

    Erstellen einer logisch-physikalischen Übersetzungstabelle (LPTT) bei der Inbetriebnahme durch Abtasten der Verwaltungsinformationen in dem Reservebereich (23) des nichtflüchtigen Speichers (11) und Speichern der logisch-physikalischen Übersetzungstabelle (LPTT) in den weiteren Speicher (12) und

    Durchführen eines logisch-physikalischen Übersetzungsprozesses auf der Grundlage der logisch-physikalischen Übersetzungstabelle des weiteren Speichers (12);

    wobei das Speicherverwaltungsverfahren durch die folgenden Schritte gekennzeichnet ist:

    Wiederherstellen, bei der Inbetriebnahme, einer Zustandskarte physikalischer Übersetzungseinheiten (STMP) zum Verwalten der physikalischen Übersetzungseinheit (TU) parallel mit dem Erstellen der logisch-physikalischen Übersetzungstabelle (LPTT) durch Abtasten der Verwaltungsinformationen in dem Reservebereich (23) des nichtflüchtigen Speichers (11); wobei

    die Zustandskarte physikalischer Übersetzungseinheiten (STMP) angibt, ob die physikalische Übersetzungseinheit (TU) in einem nicht geschriebenen sauberen Zustand CLEAN nach einer Löschoperation, einem In-Gebrauch-Zustand INUSE, in dem gültige Daten, referenziert von der logisch-physikalischen Übersetzungstabelle, gespeichert sind, oder einem ungültigen Zustand INVALID, in dem ungültige, nicht von der logisch-physikalischen Übersetzungstabelle referenzierte Daten gespeichert sind, ist, und

    während des Wiederherstellens der logisch-physikalischen Übersetzungstabelle (LPTT) und der Zustandskarte physikalischer Übersetzungseinheiten (STMP), wenn eine Vielzahl physikalischer Übersetzungseinheiten (TU) mit der gleichen logischen TU-Nummer nach Abtasten der Verwaltungsinformationen in dem Reservebereich des nichtflüchtigen Speichers existiert, Bestimmen, aus der laufenden Nummer (SQN) und der Position innerhalb des Blocks, der zuletzt geschriebenen physikalischen TU als INUSE und Bestimmen anderer physikalischer Übersetzungseinheiten TUs als INVALID.


     
    11. Programm zum Bewirken, dass ein Computer sämtliche der Schritte eines Speicherverwaltungsverfahrens nach Anspruch 10 durchführt, wenn das Programm auf dem Computer ausgeführt wird.
     


    Revendications

    1. Dispositif de mémoire (10), comprenant :

    une mémoire non volatile (11) qui permet d'écrire, de lire et d'effacer électriquement des données, et dans laquelle l'écriture et la lecture sont effectuées en unités d'une page (PAGE) et l'effacement est effectué en unités d'un bloc (BLK) comprenant une pluralité de pages (PAGE) ; et

    une section de commande (13) qui est configurée pour gérer l'accès à la mémoire non volatile (11), où la section de commande (13) est configurée pour effectuer une gestion de l'accès à la mémoire non volatile (11) en effectuant une traduction adresse logique-adresse physique (traduction logique-physique) dans des unités de traduction (TU), chacune étant une fraction entière d'une taille du bloc (BLK) et un multiple entier d'une taille de page,

    le dispositif de mémoire comprenant une mémoire supplémentaire (12), dans laquelle :

    une unité de traduction physique, TU, dans la mémoire non volatile (11) comprend une zone de données (21) et une zone de réserve (23) ; et

    la section de commande (13) est configurée :

    pour écrire des informations de gestion comprenant un numéro d'unité de traduction logique (LTUN) correspondant, et un numéro d'ordre (SQN) indiquant l'ordre d'écriture de chaque bloc (BLKm, BLKm+1, BLKm+2, BLKm+3), dans la zone de réserve (23) de la mémoire non volatile (11),

    pour construire une table de traduction logique-physique (LPTT) au démarrage en balayant les informations de gestion dans la zone de réserve (23) de la mémoire non volatile (11), et pour stocker la table de traduction logique-physique (LPTT) dans la mémoire supplémentaire (12), et

    pour exécuter un processus de traduction logique-physique sur la base de la table de traduction logique-physique de la mémoire supplémentaire (12) ;

    le dispositif de mémoire étant caractérisé en ce que :

    la section de commande (13) est configurée pour restaurer, au démarrage, une carte d'état d'unité de traduction physique (STMP) pour gérer l'unité de traduction physique (TU), parallèlement à la construction de la table de traduction logique-physique (LPTT) en balayant les informations de gestion dans la zone de réserve (23) de la mémoire non volatile (11) ; et

    la carte d'état de l'unité de traduction physique (STMP) indique si l'unité de traduction physique (TU) est dans un état non écrit, CLEAN, après une opération d'effacement, un état INUSE dans lequel des données valides référencées à partir de la table de traduction logique-physique sont stockées, ou un état INVALID dans lequel des données non valides non référencées à partir de la table de traduction logique-physique sont stockées, où :

    au cours de la restauration de la table de traduction logique-physique (LPTT) et de la carte d'état de l'unité de traduction physique (STMP), si une pluralité d'unités de traduction physique (TU) ayant le même numéro TU logique existent lors du balayage des informations de gestion dans la zone de réserve de la mémoire non volatile, la section de commande détermine, à partir du numéro d'ordre (SQN) et de la position dans le bloc, l'unité physique TU la plus récemment écrite comme étant INUSE, et détermine les autres unités de traduction (TU) physiques comme étant INVALID.


     
    2. Dispositif de mémoire selon la revendication 1, dans lequel :

    la section de commande (13) est configurée pour exécuter une opération de copie de contenus d'un bloc source d'uniquement toutes les unités de traduction physiques (TU) qui sont actuellement utilisées dans un bloc avec des unités de traduction physiques non utilisées (TU), et d'effacement du bloc source pour augmenter les unités de traduction physiques inutilisées (TU) d'un nombre égal au nombre d'unités de traduction physiques non valides (TU) incluses dans le bloc source.


     
    3. Dispositif de mémoire selon la revendication 2, dans lequel :

    lors de l'écriture d'un numéro d'unité de traduction logique (TU) dans la zone de réserve, la section de commande effectue une écriture sur une unité de traduction physique (TU) dans un bloc ayant le dernier numéro d'ordre (SQN), ladite unité de traduction physique étant l'unité TU physique CLEAN dont le numéro est le plus bas depuis sa dernière unité de traduction physique (TU).


     
    4. Dispositif de mémoire selon la revendication 3, dans lequel :

    si la dernière unité de traduction physique (TU) n'est pas CLEAN, la section de commande sélectionne de façon aléatoire un bloc qui n'a pas encore été écrit, incrémente le numéro d'ordre d'un pas, et écrit le numéro d'ordre dans le bloc.


     
    5. Dispositif de mémoire selon la revendication 4, dans lequel :

    la section de commande (13) est configurée pour mettre à jour la carte d'état de TU physique (STMP) en définissant un état d'une TU physique correspondant précédemment à une TU logique à INVALID, et en définissant un état de TU physique cible d'écriture à INUSE, et pour mettre à jour la table de traduction logique-physique (LPTT) en réécrivant sur la TU physique cible d'écriture.


     
    6. Dispositif de mémoire selon la revendication 1, dans lequel
    la section de commande (11) est configurée pour déterminer si les unités de traduction physiques (TUs) sont nouvelles ou anciennes au moyen d'une comparaison de taille des numéros d'ordre (SQN) si les TUs physiques sont présentes dans différents blocs, et sur la base de numéros de TU physiques si les TUs physiques sont présentes dans le même bloc.
     
    7. Dispositif de mémoire selon la revendication 1, dans lequel :

    lors de l'exécution d'une opération de pliage qui est une opération de copie de contenus d'un bloc source d'uniquement toutes les unités de traduction (TUs) physiques qui sont actuellement utilisées dans un bloc avec des unités de traduction physiques non utilisées (TUs), et d'effacement du bloc source pour augmenter les unités de traduction physiques inutilisées (TUs) d'un nombre égal au nombre d'unités de traduction physiques non valides (TUs) incluses dans le bloc source, la section de commande lit, par rapport à chaque TU INUSE dans un bloc cible plié, des données et des informations de gestion dans une unité de traduction physique (TU), et exécute une écriture d'une unité de traduction logique (TU) en spécifiant les données de lecture et un numéro d'unité de traduction (TU) logique, puis efface le bloc cible plié.


     
    8. Dispositif de mémoire selon la revendication 1, dans lequel :

    la section de commande (13) est configurée pour écrire des informations d'état d'achèvement d'écriture, qui permettent de déterminer si oui ou non l'écriture est terminée, dans la zone de réserve en plus des informations de gestion, et

    déterminer si oui ou non une coupure d'alimentation électrique s'est produite lors de l'écriture, en vérifiant les informations d'état d'achèvement d'écriture au démarrage par rapport au dernier bloc ayant été écrit ayant le dernier numéro d'ordre.


     
    9. Dispositif de mémoire selon la revendication 1, dans lequel :

    la section de commande (13) est configurée pour écrire des informations indiquant si oui ou non les informations de gestion sont correctes dans la zone de réserve en plus des informations de gestion, et

    vérifier si oui ou non une valeur des informations de gestion est correcte en vérifiant, au démarrage, les informations indiquant si oui ou non les informations de gestion sont correctes.


     
    10. Procédé de gestion de mémoire pour un dispositif de mémoire (10) ayant une mémoire non volatile (11) qui permet d'écrire, de lire et d'effacer électriquement des données, et où l'écriture et la lecture sont effectuées en unités d'une page (PAGE) et l'effacement est effectué en unités d'un bloc (BLK) comprenant une pluralité de pages (PAGEs), comprenant les étapes suivantes :

    effectuer une traduction adresse logique-adresse physique (traduction logique-physique) dans des unités de traduction (TUs), chacune étant une fraction entière d'une taille du bloc et un multiple entier d'une taille de page ; et

    gérer l'accès à la mémoire non volatile (11) conformément à un résultat de la traduction logique-physique,

    où le dispositif de mémoire comprend en outre une mémoire supplémentaire (12), et où

    une unité de traduction physique, TU, dans la mémoire non volatile (11) comprend une zone de données (21) et une zone de réserve (23) ; et

    le procédé comprend les étapes suivantes :

    écrire des informations de gestion comprenant un numéro de TU logique correspondant, et un numéro d'ordre (SQN) indiquant l'ordre d'écriture de chaque bloc, dans la zone de réserve de la mémoire non volatile (11),

    construire une table de traduction logique-physique (LPTT) au démarrage en balayant les informations de gestion dans la zone de réserve (23) de la mémoire non volatile (11), et stocker la table de traduction logique-physique (LPTT) dans la mémoire supplémentaire (12), et

    exécuter un processus de traduction logique-physique sur la base de la table de traduction logique-physique de la mémoire supplémentaire (12),

    le procédé de gestion de mémoire étant caractérisé par les étapes suivantes :

    restaurer, au démarrage, une carte d'état d'unité de traduction physique (STMP) pour gérer l'unité de traduction physique (TU), parallèlement à la construction de la table de traduction logique-physique (LPTT) en balayant les informations de gestion dans la zone de réserve (23) de la mémoire non volatile (11) ; où

    la carte d'état de l'unité de traduction physique (STMP) indique si l'unité de traduction physique (TU) est dans un état non écrit, CLEAN, après une opération d'effacement, un état INUSE dans lequel des données valides référencées à partir de la table de traduction logique-physique sont stockées, ou un état INVALID dans lequel des données non valides non référencées à partir de la table de traduction logique-physique sont stockées, et

    au cours de la restauration de la table de traduction logique-physique (LPTT) et de la carte d'état de l'unité de traduction physique (STMP), si une pluralité d'unités de traduction physiques (TU) ayant le même numéro TU logique existent lors du balayage des informations de gestion dans la zone de réserve de la mémoire non volatile, déterminer, à partir du numéro d'ordre (SQN) et de la position dans le bloc, l'unité physique TUs la plus récemment écrite comme étant INUSE, et déterminer les autres unités de traduction physiques comme étant INVALID.


     
    11. Programme destiné à amener un ordinateur à exécuter toutes les étapes d'un procédé de gestion de mémoire selon la revendication 10 lorsque le programme est exécuté sur l'ordinateur.
     




    Drawing

































































    Cited references

    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description