This invention relates generally to file systems for semiconductor memories.
Semiconductor memories, such as flash memories, include flash file systems that maintain a series of tables for each directory. The tables contain the contents of each directory, including files and other directories that are updated when the files and directories are added to and removed from the tables.
Reads and writes to a flash memory array tend to be relatively slow processes. Thus, each time data is accessed from the array, both the array and a file system may be accessed. Thus, to access the file, the file system must be accessed, the file located, and then the file accessed. This involves multiple reads which, over time, tend to be cumbersome and inefficient, especially for deep paths and for those directories that contain a large number of files.
Referring to
The system architecture may include a processor 510. The processor 510 may be one or more microprocessors in some embodiments of the present invention. The processor 510 is coupled by a bus 550 to other components such as a static random access memory (SRAM) 560, a wireless interface 540, and an input/output device 520. The wireless interface 540 may be a device that facilitates wireless communications over a radio frequency link in some embodiments. The wireless interface 540 may, for example, include a dipole antenna. In some embodiments, the wireless interface may facilitate cellular communications. The input/output device 520 may be any conventional input/output device including a display, a keyboard, a mouse, a keypad, or a touch screen, to mention a few examples.
In one embodiment of the present invention, the system 500 is an embedded multimedia system for multimedia applications such as playing video or audio. In some embodiments, the system 500 may be mobile and may be powered by a battery 580. However, the invention is not limited to any particular application and is equally applicable to wired, wireless, mobile, and fixed applications.
Coupled to the bus 550 is a flash memory 10. The flash memory 10 may include a bus 14 that couples a controller 12. The memory 10 may, for example, be a NOR flash memory. The controller 12 may be an embedded controller, in one embodiment. It may be a microprocessor, or some other controller in some embodiments of the present invention. The bus 14 may couple a separate random access memory 16. The memory 16 may store a file lookup table 20, a dynamic link table 22, and link list elements 24. These constituents may implement a virtual directory system. Also stored on the random access memory 16 may be software 30 and 50. The software 30 and 50 may be used to implement a virtual directory. Also coupled to the bus 14 is a flash array 18 which may store files and virtual directories, among other things.
One or more directories 26d1 and 26d2 may be stored in the file lookup table 20. The directories 26d1 and 26d2 may be handled in the same fashion as files 26f1 and 26f2. They may include the same organization, except that their lower rightmost field, as depicted in
The dynamic link table 22 is pointed to by the link table index field associated with each file 26f and each directory 26d. For example, the directory 26d1 has a link table index zero which points to a link list pointer zero in the dynamic link table 22, as indicated by the arrow extending from the file lookup table 20, and the directory 26d1, to the link list header pointer zero in the dynamic link table 22. Similarly, the directory 26d2 points to the link list header point number one, the file 26f1 points to the link list header pointer two, and the file 26f2 points to the link list header pointer three.
Each of the pointers in the dynamic link table 22 then point to a link list element 24. In some embodiments not depicted, the file lookup table 20 may point directly to the link list elements 24, eliminating the dynamic link table 22. For example, in the depicted embodiments, the link list header pointer zero points to the link list elements 24a, which includes a flash pointer, an instance, a fragment size, and a next link pointer field. In the case of the link list element 24a pointed to by the link list header zero, no further elements are pointed to. Similarly, the link list header pointer one points to the link list element 24b. The link list head pointer two points to the link list file element 24c which, in turn, points to the link list element 24d. The link list head pointer three points to the link list element 24e which, in turn, points to the element 24f, which may point to any number of additional elements, including the element 24g.
The file lookup table 20 caches file name hash values for each file and directory in the managed flash volume. It caches identifiers and parent identifiers for each file and directory in the volume. Finally, the file lookup table 20 maintains a type for each item that denotes whether it is a file or directory.
The dynamic link table 22 maintains a pointer to the head element for a specific file or directory's link list elements 24. The link list elements 24 cache a pointer to the physical data address and flash.
The link list elements 24 maintain a size of the fragment/metadata structure for efficient calculations. The size information may be used to locate file metadata, such as the file name, for comparison to the file name hash value. The size information can also be used, for example, in power on recovery to find the full sized file. The metadata may include file data or data pointers, and may be located at the head of a link list element 24.
The elements 24 may also include a field that maintains an instance number to denote a fragmentation sequence. In one embodiment, files and directories may be fragmented and stored more efficiently in non-contiguous regions within the array 18. The fragmentation sequence enables the fragments to be located and re-assembled from its fragments.
Finally, the element 24 may include a pointer to the next linked element. Thus, as illustrated in
Referring to
A free pool of space in block N 31 may be provided over the file data 23. Included within the file data is the file 26f2, the file 26f1, the directory 26d1, and the directory 26d2.
As shown in
The virtual directory system may be set up using the software 30, shown in
Initially, the file lookup table 22 is scanned to check for any instance of a new directory of the same file name with the same parent (diamond 33). If such a file exists, the flow ends. Otherwise, a header and metadata pair are written to the flash memory array 18 as indicated in block 34. The headers 27 may be written in a list, one after another, while the metadata in the region 22 may be written from the bottom up, in one embodiment. Thus, the uppermost header (with header ID O) is the header for the lowermost metadata (File name =temp). Together, the header and the metadata form a pair. The header may include an identifier (e.g., header ID O) and a fragmentation instance (e.g., instance O), a fragmentation size (e.g., 96 bytes) and a status (e.g., valid).
Then, the directory name hash from metadata is cached and stored into the file lookup table 20 as indicated in block 36. The parent ID field is set and the type is set to directory or d, as also indicated in block 36. A check at diamond 38 determines whether there is an available entry within the file lookup table 20 for this new item. Of course, as another alternative, data may be written first into the random access memory 16 and thereafter in the flash array 18.
If there is an available file lookup table 20 entry, a new link list element 24 is allocated for the header/metadata pair written to the flash memory array 18 as indicated in block 40. Then, a check at diamond 42 determines whether there is available RAM 16 that can be allocated for this entry. If so, the additional information from the flash header 27, in the array 18, is added to the newly created element 24 as indicated in block 44. The data needed to fill the fields in link list element 24 for the size of fragment and instance can be obtained from the header. The flash pointer and next link pointer can be calculated from a known offset in a block for each header and known start location for data. An entry is added to the dynamic link table 22 and the entry is set to point to the newly allocated element as indicated in block 46. Finally, the link table index of the file lookup table 20 is set to the index of the newly added dynamic link table 22 entry as indicated in block 48.
To store a file in the virtual directory, the software 50 (
As before, the file lookup table 20 is scanned, as indicated in diamond 52, to determine whether the table includes a file with the same name and parent already exists. If not, the header and metadata pair are written to the flash memory array 18 as indicated in block 54. The file name hash from metadata is cached and stored in the file lookup table 20. The parent ID field is set and the type is set to f as indicated in block 56.
A check at diamond 58 determines whether there are available entries within the table 20. If so, a new link list element 24 is allocated for the header/metadata pair written to the flash memory array 18 as indicated in block 60.
A check at diamond 62 determines whether there is available RAM space in the random access memory 16 as determined in diamond 62. If so, the information from the flash header is added to the newly created element as indicated in block 64. Then, in block 66, an entry is added to the dynamic link table 22 and the entry is set to point to the newly allocated element.
The link table index is set to the index of the newly created dynamic link table entry (block 68). This flow continues until no more data exists. The header/file data pairs are written to the flash array 18 until no more space on the block or no more file data is available as indicated in block 70. Then, the new link list element is allocated for the header/file data pair written to flash memory array 18 as indicated in block 72.
The previous link list element's next link pointer is set to the newly created element in block 76. Finally, the information from the fragment header in the flash memory array 18 is added to the newly created element as indicated in block 78.
In some embodiments of the present invention, the overhead of reads and writes to the flash memory array 18 may be reduced by creating virtual directories and files. The creation of a directory using a real directory implementation would involve the creation of data items that are generally cumbersome and inefficient for deep paths and those directories that contain a large number of files. Thus, in some embodiments, by directory virtualization, the standard deviation for directory creation time may be removed, which may be a large performance advantage over other directory implementations. Also, a virtual directory implementation based around zero byte files allows for a simple and efficient low level flash driver which can assume several key details about the data written. Thus, in some embodiments, directories may be treated as zero byte files.
By excluding the reading and writing to directory tables in flash, performance may be improved. More effective directory traversal may be achieved, in some embodiments, due to random access memory structures that contain critical file metadata. Significantly less system overhead may be used, in some embodiments, compared to other flash file systems. There may be fewer limitations in the usage model when compared to traditional flash file systems, in some embodiments, and the system may be fully power loss recoverable. In some embodiments, the data that forms the files and the directories is persistent upon asynchronous power loss once written to the managed flash area.
While the present invention has been described with respect to a limited number of embodiments, those skilled in the art will appreciate numerous modifications and variations therefrom. It is intended that the appended claims cover all such modifications and variations as fall within the true spirit and scope of this present invention.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5404485 | Ban | Apr 1995 | A |
| 5544356 | Robinson et al. | Aug 1996 | A |
| 5566314 | DeMarco et al. | Oct 1996 | A |
| 5592669 | Robinson et al. | Jan 1997 | A |
| 5630093 | Holzhammer et al. | May 1997 | A |
| 5754817 | Wells et al. | May 1998 | A |
| 5787445 | Daberko | Jul 1998 | A |
| 5822562 | Dahlen et al. | Oct 1998 | A |
| 5978808 | Wells et al. | Nov 1999 | A |
| 5987478 | See et al. | Nov 1999 | A |
| 6075858 | Schwartzman | Jun 2000 | A |
| 6167338 | De Wille et al. | Dec 2000 | A |
| 6275421 | Chen et al. | Aug 2001 | B1 |
| 6647496 | Tagawa et al. | Nov 2003 | B1 |
| 6732222 | Garritsen et al. | May 2004 | B1 |
| 6742078 | Chien et al. | May 2004 | B1 |
| 6928511 | Chiu | Aug 2005 | B2 |
| 6976133 | Wynn et al. | Dec 2005 | B1 |
| 7057942 | Suda et al. | Jun 2006 | B2 |
| 7124302 | Ginter et al. | Oct 2006 | B2 |
| 7356641 | Venkiteswaran | Apr 2008 | B2 |
| 7373521 | Kawahara | May 2008 | B1 |
| 7386700 | Lasser | Jun 2008 | B2 |
| 7426623 | Lasser | Sep 2008 | B2 |
| 20020055942 | Reynolds | May 2002 | A1 |
| 20020161860 | Godlin et al. | Oct 2002 | A1 |
| 20030009538 | Shah et al. | Jan 2003 | A1 |
| 20030041222 | Akey et al. | Feb 2003 | A1 |
| 20030065899 | Gorobets | Apr 2003 | A1 |
| 20030200232 | Matsuoka et al. | Oct 2003 | A1 |
| 20030200452 | Tagawa et al. | Oct 2003 | A1 |
| 20040010681 | Lee | Jan 2004 | A1 |
| 20040177214 | Chiu | Sep 2004 | A1 |
| 20040193864 | Tsai et al. | Sep 2004 | A1 |
| 20040199712 | Garritsen et al. | Oct 2004 | A1 |
| 20040225844 | Thompson et al. | Nov 2004 | A1 |
| 20050065946 | Shao-Hong et al. | Mar 2005 | A1 |
| 20050083794 | Tagawa et al. | Apr 2005 | A1 |
| 20050132129 | Venkiteswaran | Jun 2005 | A1 |
| 20050192686 | Hirota et al. | Sep 2005 | A1 |
| 20050193161 | Lee et al. | Sep 2005 | A1 |
| 20050195283 | Sakuda et al. | Sep 2005 | A1 |
| 20050237566 | Sakuda et al. | Oct 2005 | A1 |
| 20060026341 | Lasser | Feb 2006 | A1 |
| 20060075395 | Lee et al. | Apr 2006 | A1 |
| 20060101408 | Kotamarthi et al. | May 2006 | A1 |
| 20060161726 | Lasser | Jul 2006 | A1 |
| 20060184718 | Sinclair et al. | Aug 2006 | A1 |
| 20060224815 | Yamada et al. | Oct 2006 | A1 |
| 20060224817 | Atri | Oct 2006 | A1 |
| 20070014136 | Ali et al. | Jan 2007 | A1 |
| 20070118730 | Platt | May 2007 | A1 |
| 20080034406 | Ginter et al. | Feb 2008 | A1 |
| 20080071978 | Chow et al. | Mar 2008 | A1 |
| 20080228300 | Tagawa et al. | Sep 2008 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20060253484 A1 | Nov 2006 | US |