CN102662849A

CN102662849A - Method for managing blocks of multichannel flash memory

Info

Publication number: CN102662849A
Application number: CN2012100580811A
Authority: CN
Inventors: 颜巍; 吴斌
Original assignee: Memoright Memoritech Wuhan Co Ltd
Current assignee: Zhiyu Technology Co ltd
Priority date: 2012-03-07
Filing date: 2012-03-07
Publication date: 2012-09-12
Anticipated expiration: 2032-03-07
Also published as: CN102662849B

Abstract

The invention discloses a method for managing blocks of a multichannel flash memory. The flash memory includes several blocks of which each is divided into M physical pages where former P continuous physical pages are mapped to P logical pages in a page mapping table one by one and later Q continuous physical pages are used as reserved pages. A page writing pointer is set to each block for pointing to a blank page in the block and is added by one when every time one page is written. When page-writing operation is carried out in a target block, if one physical page mapped by one of the P logic pages is blank, then data is written on the physical page; if not, then data is written on a physical page to which the page-writing pointer points, the original physical page is marked as a garbage page, and then an address of the physical page where data is written is mapped to the original physical page. A unit of data-writing operation in the invention is reduced from a whole block to a page, which avoids a lot of erasing operation, greatly enhances an efficiency of data writing and increases a life of the flash memory.

Description

The block management method of multi-channel flash memory

Technical field

The present invention relates to a kind of block management method of multi-channel flash memory.

Background technology

For effective bandwidth and the read-write efficiency that improves flash memory, existing multi-channel flash memory generally is divided into some (block), and each piece is divided into several pages or leaves (page), and each page or leaf comprises several sectors (sector).During the multi-channel flash memory read-write, be unit-addressed with each page or leaf; But, then be to be unit with the piece for the write operation of each passage flash memory.A kind of block management method of flash memory that like Chinese patent number has been 200710072980 disclosure of the Invention; Said flash memory comprises several blocks; Every is made up of a plurality of page or leaf; Every page is divided into data storage area and redundant area, in said block, delimits one or more as the address mapping table reserved area, is that unit preserves the logical address of flash memory block and the mapping table of physical address with the page or leaf; A selected blank block is piece in return, sets up the map addresses list item of this swap block in the redundant area of said mapping table preservation page or leaf; When carrying out write operation; According to address mapping table information; The legacy data of new data and/or operated target block is write said swap block accordingly; Wipe target block, then swap block logical address physical address corresponding in the mapping table page or leaf and target block logical address physical address corresponding are exchanged.This invention can effectively prolong flash memory serviceable life and the security that improves flash data when guaranteeing operating speed.

But for the application scenario of frequent small data quantity write operation, such scheme causes block frequently erasable of flash memory, has limited the flash memory device operating efficiency, diminishes the serviceable life of flash memory.

Summary of the invention

Technical matters to be solved by this invention provides a kind of block management method of new multi-channel flash memory, further improves the flash memory device operating efficiency, prolongs the serviceable life of flash memory.

For solving the problems of the technologies described above, the present invention provides a kind of block management method of new multi-channel flash memory, and said flash memory comprises several pieces; Each piece is divided into M Physical Page; It is characterized in that in the said M Physical Page, a preceding P continuous Physical Page and the logical page (LPAGE) of the P in the page map table shines upon one by one; Back Q continuous Physical Page is as keeping page or leaf, M=P+Q; Each piece is established a page or leaf write pointer, is used in reference to blank page in piece; When all Physical Page in the piece were blank, said page or leaf write pointer value was 1; Whenever write a page or leaf, said page or leaf write pointer adds 1.

Page or leaf in carrying out object block is during write operation, in P logical page (LPAGE) of object block, chooses a logical page (LPAGE), and this chooses a Physical Page of logical page (LPAGE) mapping, is sky if choose Physical Page, then in this Physical Page write data; If choose Physical Page to have data, non-NULL, then data are write the Physical Page of said page or leaf write pointers point, then the said Physical Page of choosing is labeled as the rubbish page or leaf, the physical page address that will write data again is mapped to the said logical page (LPAGE) of choosing.

The scheme of further optimizing when the said page or leaf write operation that carries out in the object block, judges that at first whether said page or leaf write pointer value is greater than M; If said page or leaf write pointer value is greater than M; That is, when the reservation page or leaf in the object block is all write, obtain a swap block; With the non-rubbish page or leaf in the object block, be that the valid data page or leaf duplicates one by one in P Physical Page of said swap block; Then, swap block logical address physical address corresponding in the mapping table and former object block logical address physical address corresponding are exchanged, more former object block is emptied.At this moment, said swap block becomes data block, and its page write pointer value is P+1, and former object block becomes swap block.Carry out data writing operation according to new address mapping table and page or leaf write pointer then.

If said page or leaf write pointer value is not more than M, then directly carry out data writing operation according to address mapping table and page or leaf write pointer.

Beneficial effect: when the Host end will be write a page data, this method was directly programmed to page or leaf, rather than writes whole data.When the Host end will repeat to write to same logical page (LPAGE), be mapped to different Physical Page through page map table, thereby realize the page or leaf programming, write whole data again and need not wipe this piece.Because the unit of data write operation tapers to page or leaf from monoblock, has avoided a large amount of erase operations, thereby has improved the efficient of write data greatly, has improved the life-span of flash memory.

Description of drawings

Below in conjunction with accompanying drawing and embodiment technical scheme of the present invention is further specified.

Fig. 1 is the logical page (LPAGE) of a piece and the mapping synoptic diagram of Physical Page.

Fig. 2 when writing the N page operations, the mapping map of logical page (LPAGE), Physical Page.

Fig. 3 is for when page or leaf write pointer during greater than M, the mapping map of data block, swap block exchange.

Fig. 4 is a flash reading and writing process flow diagram of the present invention.

Embodiment

As shown in Figure 1, the addresses map table on the left side comprises 0～47 totally 48 logical page (LPAGE)s, and the piece table on the right comprises 0～63 M=64 Physical Page altogether, and the preceding P=48 of 64 Physical Page shines upon with 48 logical page (LPAGE)s one by one, and a back Q=16 Physical Page is as reservation page.

Each piece is established a page or leaf write pointer, is used in reference to blank page in piece; When all Physical Page in the piece were blank, a page or leaf write pointer value was 1, promptly points to the 0th Physical Page; Whenever write a page or leaf, the page or leaf write pointer adds 1.

As shown in Figure 2, when certain Physical Page carries out write operation in object block, at first,, in 48 logical page (LPAGE)s of object block, choose the N logical page (LPAGE) through the logical page (LPAGE) addressing, the N logical page (LPAGE) is mapped to the N Physical Page.At this moment, if the N Physical Page has existed data, non-NULL, then data write the 48th Physical Page that being positioned at of page write pointers point keeps page table, then the N Physical Page are labeled as the rubbish page or leaf, again with the map addresses of current the 48th Physical Page to the N logical page (LPAGE).

As shown in Figure 3; When through behind the page or leaf write operation repeatedly, the page or leaf write pointer value is greater than M, when promptly 16 in the object block keep page or leaf and all write; Application obtains a swap block, with the non-rubbish page or leaf in the object block, be that the valid data page or leaf duplicates one by one in P Physical Page of swap block.What specify is that those skilled in the art should learn that the valid data number of pages order in the object block is exactly P=48 at this moment.Then, swap block logical address physical address corresponding in the mapping table and former object block logical address physical address corresponding are exchanged, again object block is emptied.At this moment, swap block becomes new data block, and the page or leaf write pointer value of new data block is P+1, promptly points to the 48th Physical Page.Former object block becomes swap block.

Below in conjunction with Fig. 4 the flash reading and writing flow process that the present invention is overall is described.At first, flash memory gets into normal operating conditions, waits for receiving the read write command that main frame Host sends; After receiving the read write command that Host sends, flash controller carries out the address mapping table conversion, comprises by LBA being transformed into physical block address; Logical page address is transformed into physical page address, if be read data, then the mapping address according to address mapping table is addressed to Physical Page; Read this number of physical pages certificate, return then.If be write data, judge then whether the page or leaf write pointer exceeds threshold value, promptly whether the value of page or leaf write pointer, is then carried out according to address mapping table and page or leaf write pointer if not greater than M, writes the flash memory step to data, scheduler mapping table step.Specifically comprise: in P logical page (LPAGE) of object block, choose a logical page (LPAGE), this chooses a Physical Page of logical page (LPAGE) mapping, is empty if choose Physical Page, then in this Physical Page write data; If choose Physical Page to have data, non-NULL, then data are write the Physical Page of said page or leaf write pointers point, then the said Physical Page of choosing is labeled as the rubbish page or leaf, the physical page address that will write data again is mapped to the said logical page (LPAGE) of choosing.Referring to Fig. 2 explanation.

If; That is, when the reservation page or leaf in the object block is all write, obtain a swap block; With the non-rubbish page or leaf in the object block, be that the valid data page or leaf duplicates one by one in P Physical Page of swap block; Then, swap block logical address physical address corresponding in the mapping table and former object block logical address physical address corresponding are exchanged, more former object block is emptied.At this moment, swap block becomes data block, and its page write pointer value is P+1, and former object block becomes swap block.And then carry out according to an address mapping table and a page or leaf write pointer, writing the flash memory step to data, scheduler mapping table step is really accomplished the write operation of data.

It should be noted last that; Above embodiment is only unrestricted in order to technical scheme of the present invention to be described; Although with reference to preferred embodiment the present invention is specified, those of ordinary skill in the art should be appreciated that and can make amendment or be equal to replacement technical scheme of the present invention; And not breaking away from the spirit and the scope of technical scheme of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.

Claims

1. the block management method of a multi-channel flash memory; Said flash memory comprises several pieces, and each piece is divided into M Physical Page, it is characterized in that; In the said M Physical Page; A preceding P continuous Physical Page and the logical page (LPAGE) of the P in the page map table shines upon one by one, and back Q continuous Physical Page is as keeping page or leaf, M=P+Q; Each piece is established a page or leaf write pointer, is used in reference to blank page in piece; When all Physical Page in the piece were blank, said page or leaf write pointer value was 1; Whenever write a page or leaf, said page or leaf write pointer adds 1;

2. the block management method of multi-channel flash memory according to claim 1 is characterized in that, carries out page or leaf in the object block during write operation when said; Whether at first judge said page or leaf write pointer value greater than M, if said page or leaf write pointer value is greater than M, promptly; When the reservation page or leaf in the object block is all write; Obtain a swap block, with the non-rubbish page or leaf in the object block, be that the valid data page or leaf duplicates one by one in P Physical Page of said swap block, then; Swap block logical address physical address corresponding in the mapping table and former object block logical address physical address corresponding are exchanged, more former object block is emptied; At this moment, said swap block becomes data block, and its page write pointer value is P+1, and former object block becomes swap block; Carry out data writing operation according to new address mapping table and page or leaf write pointer then;