Patent Application
20230048550
The present invention relates to flash memory control, and more particularly, to a method for managing a memory apparatus, and an associated memory apparatus thereof.
While a host is accessing a memory apparatus (e.g. a solid state drive, SSD), the host typically sends an accessing command and at least a corresponding logical address to the memory apparatus. The controller of the memory apparatus receives the logical address and transfers the logical address into a physical address by utilizing a logical-to-physical address linking table. Thus, the controller accesses at least one physical memory element (or memory component) of the memory apparatus by utilizing the physical address. For example, the memory element can be implemented with one or more flash memory chips (which can be referred to as flash chips for simplicity).
The logical-to-physical address linking table can be built in accordance with a memory unit in the memory element. For example, the logical-to-physical address linking table can be built by blocks or by pages. When the logical-to-physical address linking table is built by blocks, the logical-to-physical address linking table can be referred to as the logical-to-physical block address linking table. When the logical-to-physical address linking table is built by pages, the logical-to-physical address linking table can be referred to as the logical-to-physical page address linking table. In addition, a logical-to-physical page address linking table comprising linking relationships about pages of a plurality of blocks (or all blocks) in the memory apparatus can be referred to as the global page address linking table.
Assume that the memory element has X physical blocks, and each physical block has Y physical pages. In a situation where the logical-to-physical address linking table is built by blocks, the associated logical-to-physical block address linking table can be built by reading a logical block address stored in a page of each physical block and recording the relationship between the physical block and the associated logical block. In order to build the logical-to-physical block address linking table, X pages respectively corresponding to the X physical blocks have to be read, where the time required for this is assumed to be x seconds.
In a situation where the logical-to-physical address linking table is built by pages, the associated global page address linking table can be built by reading a logical page address stored in each physical page of all physical blocks and recording the relationship between the physical page and the associated logical page. In order to build the global page address linking table, at least XY pages have to be read, requiring xY seconds. If a block has 1024 pages, the time required for building the global page address linking table is 1024 times the time required for building the logical-to-physical block address linking table, i.e. 1024·x seconds, which is an unacceptable processing time since the processing speed is too slow. That is, when implementing the global page address linking table in this way, the overall performance of accessing the memory apparatus is retarded. Therefore, a novel method is required for efficiently building the logical-to-physical address linking table, and related methods for managing memory apparatus operated under the novel method is required.
It is therefore an objective of the present invention to provide a method for managing a memory apparatus in order to solve the above-mentioned problem.
According to at least one preferred embodiment preferred embodiment of the present invention, a method for managing a memory apparatus comprising a plurality of non-volatile (NV) memory elements, each NV memory element comprising a plurality of physical blocks, and a volatile memory is provided. The method comprises: programming at least a physical block of a specific NV memory element of the NV memory elements, comprising: receiving a first host command from a host; obtaining a first host address and first data from the first host command; and linking the first host address to at least a first page of a physical block of the specific NV memory element and storing the first data into the physical block. Before the pages of the physical block are fully programmed, a temporary local page linking address table is stored in the volatile memory, and the temporary local page linking address table is updated each time a linking relationship between a page of the physical block and a host address is changed. When the memory apparatus is to be shut down, the temporary local page linking address table is written to the physical block of the specific NV memory element; and when the memory apparatus begins a start-up process, a global page address linking table is built by reading the local page address linking table.
According to another of the present invention, a memory apparatus is provided, comprising: a plurality of non-volatile (NV) memory elements, each NV memory element comprising a plurality of physical blocks, at least one of the plurality of physical blocks of each of the NV memory elements for storing a local page address linking table of the NV memory element; a volatile memory for storing a global page address linking table which is built from each local page address linking table of each of the plurality of NV memory elements; a transmission interface, for receiving commands from a host; and a processing unit, for programming at least a physical block of a specific NV memory element of the NV memory elements by obtaining a first host address and first data from a first host command, linking the first host address to at least a first page of a physical block of the specific at least one NV memory element and storing the first data in the first page, wherein before the pages of the physical block are fully programmed, the processing unit stores a temporary local page linking address table in the volatile memory, and updates the temporary local page linking address table each time a linking relationship between a page of the physical block and a host address is changed, and when the memory apparatus is to be shut down, the processing unit writes the temporary local page linking address table to the physical block of the specific NV memory element. When the memory apparatus begins a start-up process, the processing unit builds the global page address linking table by reading the local page address linking table.
These and other objectives of the present invention will no doubt become obvious to those of ordinary skill in the art after reading the following detailed description of the preferred embodiment that is illustrated in the various figures and drawings.
Please refer to
The processing unit 110 is arranged to manage the memory apparatus 100 according to a program code (not shown in
In addition, the volatile memory 120 is utilized for storing a global page address linking table, data accessed by the host (not shown), and other required information for accessing the memory apparatus 100. The volatile memory 120 of this embodiment can be a DRAM or an SRAM. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to different variations of this embodiment, the volatile memory 120 can be other kinds of volatile memories.
According to this embodiment, the transmission interface 130 shown in
In addition, the NV memory elements 140_0, 140_1, . . . , and 140_N are utilized for storing data, where the NV memory elements 140_0, 140_1, . . . , and 140_N can be, but not limited to, NAND flash chips. The bus 150 is utilized for coupling the processing unit 110, the volatile memory 120, the transmission interface 130, and the NV memory elements 140_0, 140_1, . . . , and 140_N, and for communication thereof.
According to this embodiment, the processing unit 110 can provide at least one block of the memory apparatus 100 with at least one local page address linking table within the memory apparatus 100, where the local page address linking table comprises linking relationships between physical page addresses and logical page addresses of a plurality of pages. In this embodiment, the processing unit 110 builds the local page address linking table during programming/writing operations of the memory apparatus 100. The processing unit 110 can further build the global page address linking table mentioned above according to the local page address linking table. For example, the processing unit 110 reads a first linking relationship between a first physical page address and a first logical page address from the at least one local page address linking table, and then records the first linking relationship into the global page address linking table. The processing unit 110 can further read a second linking relationship between a second physical page address and the first logical page address from the at least one local page address linking table, and then record the second linking relationship into the global page address linking table in order to update the global page address linking table.
More particularly, the processing unit 110 provides a plurality of blocks of the memory apparatus 100 with a plurality of local page address linking tables within the memory apparatus 100, respectively. That is, the aforementioned at least one local page address linking table comprises a plurality of local page address linking tables. The processing unit 110 can further build the global page address linking table mentioned above according to the local page address linking tables. More specifically, the processing unit 110 can read one of the local page address linking tables to update the global page address linking table mentioned above. For example, the first linking relationship of a first physical page is read from a first local page address linking table of the local page address linking tables, and the second linking relationship of a second physical page is read from a second local page address linking table of the local page address linking tables. Implementation details of the local page address linking tables are further described by referring to
According to this embodiment, the page amount of the data area and the page amount of the table area can be determined as required. For example, pages 0, 1, 2, . . . , 126 is utilized for storing data and the remaining page of the block is utilized for storing the local page address linking table 0. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the data area may comprise less than 127 pages, and the table area may comprise two or more pages. According to another variation of this embodiment, the total page amount of the block, the page amount of the data area, and the page amount of the table area may vary. Please note that a page is a programming/writing unit. In other words, when programming/writing data is required, the processing unit 110 programs/writes a page of data into a page at a time. According to this embodiment, the NV memory elements 140_0, 140_1, . . . , and 140_N shown in
In this embodiment, the local page address linking table 0 is built when all the data pages in the block 0 have been programmed, namely fully programmed. Before the data pages in the block 0 are fully programmed, however, the processing unit 110 temporarily stores a temporary local page address linking table 0 in the volatile memory 120, and further updates the temporary local page address linking table 0 when any linking relationship between a physical page address and a logical page address in the block 0 is changed.
According to this embodiment, the ranking of a field (entry) of the temporary/non-temporary local page address linking table (e.g. the temporary local page address linking table 0 or the local page address linking table 0) represents a physical page address, and the content of this field represents an associated logical page address. For example, suppose that iP and jP are respectively the row number and the column number of the illustrative table location (iP, jP) of the temporary/non-temporary local page address linking table shown in
PPN=(PBN*DPC+iP*4+jP);
where the notation PBN stands for the physical block number of the physical block under discussion (e.g. PBN=0, 1, 2, . . . , etc. for the blocks 0, 1, 2, . . . , etc., respectively), and the notation DPC stands for the data page count of each block (e.g. 127 in this embodiment). This is for illustrative purposes only, and is not meant to be a limitation of the present invention. For better comprehension, the temporary/non-temporary local page address linking table can be illustrated as a single column, as shown in the right half of
PPN=(PBN*DPC+iP).
Please note that, in this embodiment, a range of the logical page addresses in the local page address linking table 0 is not greater than the number of pages in the block 0 (i.e. 128 in this embodiment). This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, a range of the logical page addresses in a local page address linking table such as the local page address linking table 0 can be greater than the number of pages in a block such as the block 0.
Within the temporary local page address linking table 0 or the local page address linking table 0 shown in
According to the embodiment shown in
Similarly, when the host then sends a command 1 to the processing unit 110 in order to program data 1 at a logical page address 0x0001, the processing unit 110 programs the data 1 and the logical page address 0x0001 into the page 1 of the block 0 of the flash chip 0, wherein the data 1 is programmed in a data byte region (labeled “DBR”) of the page 1, and the logical page address 0x0001 is programmed in a spare byte region (labeled “SBR”) of the page 1 as spare information. In addition, the processing unit 110 writes the logical page address 0x0001 into the second field of the temporary local page address linking table 0 (or the illustrative table location (0, 1) thereof in this embodiment, i.e. the illustrative table location of the second column and the first row) to thereby indicate that the logical page address 0x0001 links/maps to page 1 of block 0 of flash chip 0, whose physical page address is 0x0001. Afterward, when the host sends a command 2 to the processing unit 110 in order to program data 2 at the logical page address 0x0002 again, the processing unit 110 programs the data 2 and the logical page address 0x0002 into the page 2 of the block 0, wherein the data 2 is programmed in a data byte region (labeled “DBR”) of the page 2, and the logical page address 0x0002 is programmed in a spare byte region (labeled “SBR”) of the page 2 as spare information. In addition, the processing unit 110 writes the logical page address 0x0002 into the third field of the temporary local page address linking table 0 (or the illustrative table location (0, 2) thereof in this embodiment, i.e. the illustrative table location of the third column and the first row) to thereby update that the logical page address 0x0002 links/maps to the page 2 of the block 0 of the flash chip 0, whose physical page address is 0x0002. Similar operations for the subsequent pages are not repeated in detail for simplicity.
As a result of the above operations, referring to the upper-right portion of
In this variation, after programming a first portion of data pages of a specific block, the processing unit 110 can program a first local page address linking table for the first portion of data pages, where the first local page address linking table is positioned next to the first portion of data pages. After programming a second portion of data pages of the specific block, the processing unit 110 can program a second local page address linking table for the second portion of data pages. For example, the second local page address linking table is positioned next to the second portion of data pages. In another example, the second local page address linking table is positioned at the end (e.g. the last page) of the specific block. In another example, the second local page address linking table is positioned at the beginning (e.g. the first page) of the block next to the specific block. In another example, the second local page address linking table is positioned at another page (or other pages) of the block next to the specific block.
According to this embodiment, referring to the left half of
Within the global page address linking table shown in the left half of
When building the global page address linking table, the processing unit 110 reads the first field of the local page address linking table 0 shown in
Next, the processing unit 110 reads the second field of the local page address linking table 0 shown in
Then, the processing unit 110 reads the third field of the local page address linking table 0 shown in
Subsequently, the processing unit 110 reads the fourth field of the local page address linking table 0 shown in
Afterward, the processing unit 110 reads the fifth field of the local page address linking table 0 shown in
According to the embodiment shown in
Next, the processing unit 110 reads the second field of the local page address linking table 1 shown in
Instead of reading all pages (or memory units) of the NV memory elements 140_0, 140_1, . . . , and 140_N to build the global page address linking table, the processing unit 110 of this embodiment merely reads a few number of local page address linking tables within (or representing but not within) the blocks that are fully or partially programmed. Therefore, the memory apparatus implemented according to the present invention surely have better efficiency than those implemented according to the related art.
According to a variation of this embodiment, in a situation where all data pages of all data blocks of the NV memory elements 140_0, 140_1, . . . , and 140_N are fully programmed, the processing unit 110 merely reads the local page address linking tables respectively corresponding to the data blocks to build the global page address linking table. If the NV memory elements 140_0, 140_1, . . . , and 140_N have XD data blocks in total, and each data block has YD data pages, the processing unit 110 reads XD local page address linking tables (whose data amount is typically less than XD pages in total) to build the global page address linking table, rather than reading XD·YD pages. In other words, the time required for building the global page address linking table according to the present invention is similar to the time required for building the global block address linking table.
According to another variation of this embodiment, in a situation where a particular block is not fully programmed (i.e. the particular block is partially programmed), at one time there is no local page address linking table within the particular block. In the volatile memory 120, however, there is a temporary local page address linking table of the particular block. The processing unit 110 of this variation can program/write the temporary local page address linking table to the particular block before shutting down the memory apparatus 100. For example, after the memory apparatus 100 begins a start-up process, the host can read the local page address linking table stored in the particular block, in order to build or update the global page address linking table. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. In another example, after the memory apparatus 100 begins a start-up process, the processing unit 110 can read the pages programmed in the particular block, and more particularly, the spare byte region of each page programmed in the particular block, in order to build or update the global page address linking table.
In a situation where the processing unit 110 reads the pages programmed in the particular block to build or update the global page address linking table, the processing unit 110 has to read less than YD pages of data from the particular block. As a result, for completing the global page address linking table, the data amount that the processing unit 110 has to read is less than (XFP+YPP) pages, given that the NV memory elements 140_0, 140_1, . . . , and 140_N have XFP fully programmed blocks in total and further have a partially programmed block having YPP programmed data pages. Therefore, in regard to building the global page address linking table, the memory apparatus implemented according to the present invention still have better efficiency than those implemented according to the related art.
According to different variations of the embodiments mentioned above, the global page address linking table can be built during any start-up process of the memory apparatus 100 or at any time in response to a request from a user.
According to different variations of the embodiments mentioned above, the global page address linking table can be divided into a plurality of partial tables stored in one or more of the NV memory elements (e.g. the partial tables are respectively stored in the NV memory elements 140_0, 140_1, . . . , and 140_N). Each divided partial table can be referred as a sub-global page address linking table. The processing unit 110 can read and store at least one sub-global page address linking table (e.g. a sub-global page address linking table, some sub-global page address linking tables, or all the sub-global page address linking tables) of the global page address linking table into the volatile memory 120, depending on the size of the global page address linking table and the size of the volatile memory 120 or depending on some requirements. The processing unit 110 can utilize the sub-global page address linking table stored in the volatile memory 120 to perform the logical-to-physical address transferring operations of the aforementioned embodiments.
Regarding the physical block addresses, the first block of the flash chip 0 is regarded as the first block of the flash chips 0-3, and is addressed as the physical block address 0, and therefore, can be referred to as PHY BLK 0, where “PHY BLK” stands for “physical block”. The last block of the flash chip 0 is regarded as the 1024th block of the flash chips 0-3, and is addressed as the physical block address 1023, and therefore, can be referred to as PHY BLK 1023. The first block of the flash chip 1 is regarded as the 1025th block of the flash chips 0-3, and is addressed as the physical block address 1024, and therefore, can be referred to as PHY BLK 1024, and so on. The last block of the flash chip 3 is regarded as the 4096th block of the flash chips 0-3, and is addressed as the physical block address 4095, and therefore, can be referred to as PHY BLK 4095. In this embodiment, the blocks of the flash chips 0-3 comprise 4 sets of PHY BLKs {0, 1, . . . , 1023}, {1024, 1025, . . . , 2047}, {2048, 2049, . . . , 3071}, and {3072, 3073, . . . , 4095}, i.e. 4096 PHY BLKs in total.
Regarding the physical page addresses, the first page of PHY BLK 0 is regarded as the first page of the flash chips 0-3, and is addressed as the physical page address 0, and therefore, can be referred to as PHY Page 0. The last page of PHY BLK 0 is regarded as the 128th page of the flash chips 0-3, and is addressed as the physical page address 127, and therefore, can be referred to as PHY Page 127. The first page of PHY BLK 1 is regarded as the 129th page of the flash chips 0-3, and is addressed as the physical page address 128, and therefore, can be referred to as PHY Page 128, and so on. The last page of PHY BLK 4095 is regarded as the 524288th page of the flash chips 0-3, and is addressed as the physical page address 524287, and therefore, can be referred to as PHY Page 524287. In this embodiment, the pages of the flash chips 0-3 comprise 4096 sets of PHY Pages {0, 1, . . . , 127}, {128, 129, . . . , 255}, . . . , and {524160, 524161, . . . , 524287}, i.e. 524288 PHY Pages in total.
During writing/programming operations, the host sends a command C0 to the memory apparatus 100 in order to write 4 sectors of data, DS0-DS3, at corresponding host addresses 0000008-0000011. The volatile memory 120 temporarily stores data DS0-DS3. The processing unit 110 parses the command C0 to execute the writing/programming operation corresponding to the command C0. The processing unit 110 transfers the host addresses 0000008-0000011 into associated logical addresses. The processing unit 110 divides the host address 0000008 by the number of sectors of a page, i.e. 4 in this embodiment, and obtains a quotient 2 and a remainder 0. The quotient 2 means that the logical page address thereof is 2; therefore, the logical page indicated by the logical page address 2 can be referred to as LOG Page 2. In addition, the remainder 0 means that the data DS0 should be stored in a first sector of a page. The processing unit 110 further divides the host address 0000008 by the number of sectors of a block, i.e. 512 in this embodiment, and obtains a quotient 0 and a remainder 8. The quotient 0 means that the logical block address thereof is 0; therefore, the logical block indicated by the logical block address 0 can be referred to as LOG BLK 0, where “LOG BLK” stands for “logical block”.
In practice, when the host address is expressed with the binary numeral system, the dividing operations can be performed by truncating a portion of bits of the host address. For example, when dividing the host address 0000008 by 4, the processing unit 110 extracts the last two bits (i.e. two adjacent/continuous bits including the least significant bit (LSB)) from the binary expression of the host address to obtain the remainder 0, and extracts the other bits from this binary expression to obtain the quotient 2. In addition, when dividing the host address 0000008 by 512, the processing unit 110 can extract the last nine bits (i.e. nine adjacent/continuous bits including the LSB) from the binary expression of the host address to obtain the remainder 8, and extract the other bits from this binary expression to obtain the quotient 0. Therefore, in this embodiment, the host address 0000008 substantially comprises the logical page address 2 and the logical block address 0. Please note that, as the host address 0000008 inherently belongs to LOG Page 2 and inherently belongs to LOG BLK 0, the processing unit 110 of a variation of this embodiment can parse the host address 0000008 by bit-shifting, rather than really performing the dividing operations.
Similarly, the processing unit 110 of this embodiment determines that the logical page addresses of the host addresses 0000009, 0000010, and 0000011 are all 2 (i.e. all of the host addresses 0000009, 0000010, and 0000011 inherently belong to LOG Page 2, or comprise the logical page address 2), and the logical block addresses thereof are all 0 (i.e. all of the host addresses 0000009, 0000010, and 0000011 further inherently belong to LOG BLK 0, or comprise the logical block address 0). In addition, the data DS1, DS2, and DS3 should be respectively stored in the second, the third, and the fourth sectors of a page.
In this embodiment, PHY BLK 0 is erased and is logically positioned in the spare region initially, the processing unit 110 pops the PHY BLK 0 from the spare region, and writes/programs the data DS0-DS3 into the first, the second, the third, and the fourth sectors of PHY Page 0, respectively. The processing unit 110 further records 0 in the third field of the global page address linking table of this embodiment, in order to indicate that LOG Page 2 links to PHY Page 0.
In addition, the processing unit 110 records usage information during accessing the pages. For example, the usage information comprises a valid page count table for recording valid page counts of the blocks, respectively. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the usage information comprises an invalid page count table for recording invalid page counts of the blocks, respectively. In practice, as each fully programmed block comprise a predetermined number of pages (e.g. 128 pages in this embodiment), the valid page count and the invalid page count of the same fully programmed block are complementary to each other.
According to this embodiment, the processing unit 110 records 1 in the first field of the valid page count table, in order to indicate that PHY BLK 0 contains 1 valid page (i.e. 1 page of useful data; or in other words, 1 page of valid data). Please note that the global page address linking table and the valid page count table can be stored in the volatile memory 120. In this way, the global page address linking table and the valid page count table can be updated easily during accessing the flash chips. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the global page address linking table and the valid page count table can be loaded from the volatile memory 120 and stored in one or more of the NV memory elements 140_0, 140_1, . . . , and 140_N before shutting down the memory apparatus 100. More particularly, the global page address linking table and the valid page count table can be stored in one or more link blocks of the NV memory elements 140_0, 140_1, . . . , and 140_N. In this way, the global page address linking table and the valid page count table can be preserved while the memory apparatus 100 shuts down. Each of the one or more link blocks is a particular block for preserving system information. While turning on the memory apparatus 100 next time, the global page address linking table and the valid page count table can be easily obtained from the link block(s).
Next, the host sends a command C1 to the memory apparatus 100 in order to write 4 sectors of data, DS4-DS7, into corresponding host addresses 0000512-0000515. Similarly, the processing unit 110 determines that the logical page addresses of the host addresses 0000512-0000515 are all 128 (i.e. all of the host addresses 0000512-0000515 belong to LOG Page 128, or comprise the logical page address 128), and the logical block addresses thereof are all 1 (i.e. all of the host addresses 0000512-0000515 further belong to LOG BLK 1, or comprise the logical block address 1). In addition, the data DS4-DS7 should be stored in the first, the second, the third, and the fourth sectors of a page, respectively. Since PHY Page 0 has been programmed, the processing unit 110 writes/programs the data DS4-DS7 into the first, the second, the third, and the fourth sectors of PHY Page 1 (which is the page subsequent to PHY Page 0), respectively. The processing unit 110 further records 1 in the 129th field of the global page address linking table shown in
Please note that the host addresses 0000512-0000515 and the host addresses 0000008-0000011 belong to different logical block (e.g. the host addresses 0000512-0000515 belong to LOG BLK 1, and the host addresses 0000008-0000011 belong to LOG BLK 0), however, these host addresses all link to the associated pages in the same physical blocks, and data corresponding to the host addresses 0000512-0000515 and data corresponding to the host addresses 0000008-0000011 are both programmed/written in the same physical block, i.e. PHY BLK 0 in this embodiment.
In the above situation, when a first set of host addresses (e.g. the host addresses 0000512-0000515) belong to a first logical block (e.g. LOG BLK 1) and a second set of host addresses (e.g. the host addresses 0000008-0000011) belong to a second logical block (e.g. LOG BLK 0), the processing unit 110 can program/write both the data corresponding to the first set of host addresses and the data corresponding to the second set of host addresses in the same physical block (e.g. PHY BLK 0). This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, when a first set of host addresses belong to a first logical block, the processing unit 110 can program/write a first portion and a second portion of the data corresponding to the first set of host addresses in different physical blocks wherein the first portion and the second portion of the data are not overlap.
In this embodiment, the host then sends a command C2 to the memory apparatus 100 in order to write 4 sectors of data, DS8-DS11, into corresponding host addresses 0000004-0000007. Similarly, the processing unit 110 determines that the logical page addresses of the host addresses 0000004-0000007 are all 1 (i.e. all of the host addresses 0000004-0000007 belong to LOG Page 1, or comprise the logical page address 1), and the logical block addresses thereof are all 0 (i.e. all of the host addresses 0000004-0000007 further belong to LOG BLK 0, or comprise the logical block address 0). In addition, the data DS8-DS11 should be stored in the first, the second, the third, and the fourth sectors of a page, respectively. Since PHY Page 1 has been programmed, the processing unit 110 writes/programs the data DS8-DS11 into the first, the second, the third, and the fourth sectors of PHY page 2 (which is the page subsequent to PHY Page 1), respectively. The processing unit 110 further records 2 in the second field of the global page address linking table shown in
Subsequently, the host sends a command C3 to the memory apparatus 100 in order to write/update 4 sectors of data, DS0′-DS3′, into corresponding host addresses 0000008-00000011. Similarly, the processing unit 110 determines that the logical page addresses of the host addresses 0000008-00000011 are all 2 (i.e. all of the host addresses 0000008-0000011 belong to LOG Page 2, or comprise the logical page address 2), and the logical block addresses thereof are all 0 (i.e. all of the host addresses 0000008-0000011 further belong to LOG BLK 0, or comprise the logical block address 0). In addition, the data DS0′-DS3′ should be stored in the first, the second, the third, and the fourth sectors of a page, respectively. Since PHY Page 2 has been programmed, the processing unit 110 writes/programs the data DS0′-DS3′ into the first, the second, the third, and the fourth sectors of PHY page 3 (which is the page subsequent to PHY Page 2), respectively. The processing unit 110 further records/updates 3 in the third field of the global page address linking table shown in
Although 4 pages, PHY Pages 0-3, have been programmed in PHY BLK 0, only 3 physical pages, PHY Page 1-3, contain valid data. Since data of LOG Page 2 has been updated, PHY Page 0 does not contain valid data and can be deemed as an invalid page containing invalid data. As a result of executing command C3, the global page address linking table and the valid page count table are updated as shown in
In this embodiment, referring to
As a result of executing command C4, the global page address linking table and the valid page count table are updated as shown in
Next, the host sends a command C5 to the memory apparatus 100 in order to read 4 sectors of data corresponding host addresses 0000008-00000011. The processing unit 110 parses the command C5 to execute the reading operation. The processing unit 110 transfers the host addresses 0000008-0000011 into logical addresses. The processing unit 110 divides the host address 0000008 by the number of sectors of a page, i.e. 4 in this embodiment, and obtains a quotient 2 and a remainder 0. The quotient 2 means that the logical page address thereof is 2, where the logical page indicated by the logical page address 2 is LOG Page 2. In addition, the remainder 0 means that the data DS0 should have been stored in the first sector of a page. Similarly, the processing unit 110 determines that the logical page addresses of the host addresses 0000009, 0000010, and 0000011 are all 2 (i.e. all of the host addresses 0000009, 0000010, and 0000011 belong to LOG Page 2, or comprise the logical page address 2), and the logical block addresses thereof are all 0 (i.e. all of the host addresses 0000009, 0000010, and 0000011 further belong to LOG BLK 0, or comprise the logical block address 0). In addition, the data corresponding to host addresses 0000008-00000011 should have been stored in the first, the second, the third, and the fourth sectors of a page, respectively. The processing unit 110 reads the third field of the global page address linking table and obtains 128, which indicates that the data corresponding to LOG Page 2 is stored in PHY Page 128. The processing unit 110 reads PHY Page 128 to obtain data DS0″-DS3″, and sends these data to the host.
In this embodiment, assume that after a lot of writing/programming operations are further performed, all blocks of the data region (e.g. PHY BLKs 0-4093) have been fully programmed, and the spare region comprises PHY BLKs 4094 and 4095, where the valid page count table is illustrated in
Assuming that maintaining a sufficient block count of the spare region is required in this embodiment, in a situation where the block count of the spare region is (or will be) less than a predetermined value (e.g. the predetermined value is 2), the processing unit 110 has to erase a physical block in the data region, in order to push this erased physical block into the spare region. The processing unit 110 searches the valid page count table and finds out that PHY BLK 2 has no valid data since the valid page count of PHY BLK 2 is 0. Since, PHY BLK 2 having the least valid page count, the processing unit 110 erases the PHY BLK 2 and then pushes the erased PHY BLK 2 into the spare region. Thus, the spare region comprises PHY BLKs 2 and 4095 now. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, when the valid page count of PHY BLK 2 decreases to zero, the processing unit 110 can erase PHY BLK 2 immediately.
According to this embodiment, assume that after several writing/programming operations are further performed, all pages of the PHY BLK 4094 have been programmed, where the valid page count table is illustrated in
Similarly, when it is detected that the block count of the spare region is (or will be) less than the predetermined value, the processing unit 110 has to erase at least a physical block in the data region in order to push the physical block(s) into the spare region. The processing unit 110 of this embodiment searches the valid page count table shown in
In this embodiment, when it is detected that the block count of the spare region is less than the predetermined value, the processing unit 110 typically searches the valid page count table to find one or more fully programmed blocks having the least valid page count(s), and erases the one or more fully programmed blocks in order to push the one or more blocks into the spare region. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, in a situation where the valid page count table is replaced by the invalid page count table mentioned above, the processing unit 110 can search the invalid page count table to find one or more fully programmed blocks having the most invalid page count(s), and erase the one or more fully programmed blocks of this variation in order to push the one or more blocks into the spare region.
According to this embodiment, the processing unit 110 has popped one more physical block from the spare region into the data region, such as PHY BLK 2, for merging PHY BLKs 0 and 1. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the processing unit 110 can merge the one or more fully programmed blocks having the least valid page count(s) into a partially programmed block as long as there are enough free pages in the partially programmed block, where the free pages represent the pages that have not been programmed since the latest erasure of the block comprising these valid pages. For example, the processing unit 110 can merge PHY BLKs 0 and 1 into the partially programmed block, such as PHY BLK 4095, as long as there are enough free pages in the partially programmed block for programming data DS16-DS19 and the valid data of PHY BLKs 0 and 1. In another example, the processing unit 110 can merge PHY BLK 0 into the partially programmed block, such as PHY BLK 4095, as long as there are enough free pages in the partially programmed block for programming data DS16-DS19 and the valid data of PHY BLK 0.
In practice, the processing unit 110 can program/write the data DS16-DS19 into PHY BLK 4095, and can further move the valid data of PHY BLKs 0 and 1 into PHY BLK 4095 as long as there are enough free pages in PHY BLK 4095 for programming data DS16-DS19 and the valid data. Certainly, the processing unit 110 of this variation updates the global page address liking table to reflect the movement of the valid data. Similarly, after moving the valid data, the processing unit 110 erases PHY BLKs 0 and 1, and pushes the erased PHY BLKs 0 and 1 into the spare region.
In other variations of this embodiment, the processing unit 110 can move valid data of N physical blocks into M physical blocks wherein N and M are positive integers, and N is greater than M. Assume that there are K pages of valid data in total within the N physical blocks, where K is smaller than the number of free pages in total within the M physical blocks. The processing unit 110 can read the K pages of valid data from the N physical blocks, erase the N physical blocks, buffer the K pages of valid data into the volatile memory 120, and program/write the K pages of valid data into the M physical blocks. Please note that, in general, the N physical blocks and the M physical blocks may overlap (e.g. the N physical blocks and the M physical blocks both comprise at least a same physical block) or not overlap. In a situation where the N physical blocks and the M physical blocks do not overlap (i.e. none of the N physical blocks belongs to the M physical blocks, and vice versa), the K pages of valid data can be programmed/written into the M physical blocks without waiting for erasing the N physical blocks, and the processing unit 110 can generate (N−M) erased blocks eventually. Certainly, the processing unit 110 updates the global page address liking table to reflect the movement of the valid data.
Please note that, in other variations of this embodiment, the processing unit 110 can record the invalid page count of each physical block. For example, given that the page count of each physical block is 128, a specific physical block comprises 128 pages, within which 28 pages are invalid pages containing invalid data and 100 pages are valid pages containing valid data. That is, the invalid page count and the valid page count of the specific physical block are 28 and 100, respectively. The processing unit 110 can build an invalid page count table of the flash chips 0-3, and erase a particular physical block according to the invalid page count table. In some of the variations, when the processing unit 110 has to erase a physical block, the processing unit 110 can select a particular physical block having the most invalid pages according to the invalid page count table, and erase the particular physical block. In practice, before the particular physical is erased, the valid data contained therein have to be moved to other blocks. For efficiently moving the valid data, the processing unit 110 can record one or more positions of the valid data in the particular block. More particularly, the processing unit 110 can build a valid-page-position table for each block in order to indicate the position(s) of one or more valid pages containing valid data within the block.
In particular, each field of the valid-page-position table indicates the valid-page-position(s) corresponding to the associated physical block. Each bit in a specific field indicates whether an associated page in the associated physical block is valid or invalid. For example, the first field of the valid-page-position table shown in
More specifically, the ranking of a specific bit in the specific field of the valid-page-position table shown in
In this embodiment, the LSB in the specific field indicates whether the first page of the associated physical block is a valid page or an invalid page, and the MSB in the specific field indicates whether the last page of the associated physical block is a valid page or an invalid page. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the LSB in the specific field indicates whether the last page of the associated physical block is a valid page or an invalid page, and the MSB in the specific field indicates whether the first page of the associated physical block is a valid page or an invalid page. For example, regarding the bits “01011100101 11111” in the first field, the LSB “1” indicates that the last page of the PHY BLK 0 (i.e. the PHY Page 127) is a valid page containing valid data, and the most significant bit (MSB) “0” indicates that the first page of the PHY BLK 0 (i.e. the PHY Page 0) is an invalid page containing invalid data, where other bits between LSB and MSB indicate the valid/invalid state of the other physical pages of the associated physical block, respectively.
In this embodiment, a logical value “1” of the specific bit indicates that the associated page is a valid page, while a logical value “0” of the specific bit indicates that the associated page is an invalid page. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the logical value “0” of the specific bit indicates that the associated page is a valid page, while the logical value “1” of the specific bit indicates that the associated page is an invalid page.
In addition, the valid-page-position table can be stored in the volatile memory 120. In this way, the valid-page-position table can be updated easily during accessing the flash chips. This is for illustrative purposes only, and is not meant to be a limitation of the present invention. According to a variation of this embodiment, the valid-page-position table can be loaded from the volatile memory 120 and stored in one or more of the NV memory elements 140_0, 140_1, . . . , and 140_N before shutting down the memory apparatus 100. More particularly, the valid-page-position table can be stored in one or more link blocks of the NV memory elements 140_0, 140_1, . . . , and 140_N. In this way, the valid-page-position table can be preserved while the memory apparatus 100 shuts down. While turning on the memory apparatus 100 next time, the valid-page-position table can be easily obtained from the link block(s).
In another embodiment, during accessing the memory apparatus 100, the valid-page-position table and global page address linking table can be loaded from the volatile memory 120 and stored in the NV memory elements from time to time. For example, the valid-page-position table and global page address linking table can be stored in every predetermined time period (e.g. 2 second) or in every predetermined accessing operations (e.g. 100 times of writing). When the memory apparatus 100 is abnormally shutting down, the latest valid-page-position table and global page address linking table are not loaded from the volatile memory 120 and stored in the NV memory elements. Then, the memory apparatus 100 is turning on. For building the valid-page-position table, the processing unit 110 can search the blocks that have been accessed after the latest updating of the valid-page-position table and global page address linking table in the NV memory elements. The processing unit 110 searches logical page addresses stored in each page of these blocks to build and update the global page address linking table. After that, the valid-page-position table can be built according to the updated global page address linking table.
In contrast to the related art, the present invention method and apparatus can greatly save the time of building logical-to-physical page address linking table(s), such as the global page address linking table. Therefore, the present invention provides better performance than the related art.
It is another advantage of the present invention that the present invention method and apparatus can record the usage information during accessing the pages, and therefore can efficiently manage the usage of all blocks according the usage information. As a result, the arrangement of the spare region and the data region can be optimized.
In addition, managing the flash memory on a basis of page brings lots of advantages. For example, the speed of random write is greatly improved, and the write amplification index can be greatly reduced. Without introducing side effects such as those of the related art, managing the flash memory on a basis of page can be much simpler and more intuitional than managing the flash memory on a basis of block as long as the present invention is applied in real implementation.
Those skilled in the art will readily observe that numerous modifications and alterations of the device and method may be made while retaining the teachings of the invention. Accordingly, the above disclosure should be construed as limited only by the metes and bounds of the appended claims.
This application is a continuation application of U.S. application Ser. No. 17/351,168, filed on Jun. 17, 2021, which is a continuation application of U.S. application Ser. No. 16/888,836, filed on May 31, 2020, which is a continuation application of U.S. application Ser. No. 16/596,703, filed on Oct. 8, 2019, which is a continuation application of U.S. application Ser. No. 15/642,295, filed on Jul. 5, 2017, which is a continuation application of U.S. application Ser. No. 14/566,724, filed on Dec. 11, 2014, which is a continuation application of U.S. application Ser. No. 13/604,644, filed on Sep. 6, 2012, which is a continuation application of U.S. application Ser. No. 12/471,462, filed on May 25, 2009, which claims the benefit of U.S. Provisional Application No. 61/112,173, filed on Nov. 6, 2008, and claims the benefit of U.S. Provisional Application No. 61/140,850, filed on Dec. 24, 2008. The contents of these applications are incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 61112173 | Nov 2008 | US | |
| 61140850 | Dec 2008 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 17351168 | Jun 2021 | US |
| Child | 17975565 | US | |
| Parent | 16888836 | May 2020 | US |
| Child | 17351168 | US | |
| Parent | 16596703 | Oct 2019 | US |
| Child | 16888836 | US | |
| Parent | 15642295 | Jul 2017 | US |
| Child | 16596703 | US | |
| Parent | 14566724 | Dec 2014 | US |
| Child | 15642295 | US | |
| Parent | 13604644 | Sep 2012 | US |
| Child | 14566724 | US | |
| Parent | 12471462 | May 2009 | US |
| Child | 13604644 | US |
