Patent Application
20070005929
The present invention relates to flash memory devices and more specifically to flash memory filesystems.
Currently, flash memory requires specially designed flash filesystems. This is because flash memory devices are unable to rewrite arbitrary data, and also because flash memory contains large erase blocks, which makes erasing and rewriting an entire region too time consuming to be practical.
Creation and distribution of such filesystems is not feasible in all applications. For example, a flash memory device in a card form factor may be used as a removable storage media. The flash device is ideally available to an operating system as a set of arbitrarily rewritable sectors or slots, much like a hard drive.
Random access memory (RAM) may be included in the flash device to store a locator table, thus allowing the device to be accessed in the same manner as a hard drive. However, the required amount of RAM to store the locator table may be quite large and thus cost prohibitive.
A better understanding of the present invention can be obtained from the following detailed description in conjunction with the following drawings, in which:
In the following description, for purposes of explanation, numerous details are set forth in order to provide a thorough understanding of embodiments of the present invention. However, it will be apparent to one skilled in the art that these specific details are not required in order to practice the present invention as hereinafter claimed.
As used herein, a slot refers to a physical region within a data block. A sector refers to the logical address that indexes into a mapping table and its chains. The sector resolves to a logical block number and to a slot within that logical data block. For example, when a drive is plugged into a computer, a certain capacity for the drive is reported; the operating system may write and read sectors using a FAT filesystem. In embodiments of the present invention, sectors may be translated to slots in data blocks using mapping tables.
Data blocks (110) contain slots (116) which may be programmed with data, read, or erased. A slot may be any size, but is typically much smaller than a block. The slots within each data block may include allocated slots as well as non-reclaimed dirty slots. Slot status bits (114) within the data blocks indicate whether each slot within the data block is free, allocated, or dirty (deprecated). The data blocks (110) also include a header (112). The header (112) may contain information including the logical block number associated with the data block, power loss recovery (PLR) state information, or other meta-information.
Mapping blocks (120) contain a table or array (124) which locates a given slot in the data block region (110). The mapping table (124) is organized using a link chain and a locator or set of locators per record, including a terminal locator. The locators are used to determine the physical address of a slot within the flash memory array by using a logical block number and slot within the block. Using more than one locator per record increases efficiency of space usage. For example, one link with one locator is 50% efficient. One link having three locators is 75% efficient. The number of locators per link may be optimized by analyzing the number of times a slot is written before reclaim. The goal of optimization is to ensure that the majority of sectors do not require the use of overflow records (134), described below, and to increase runtime efficiency by skipping invalid locator records each time a link is followed.
Each mapping block (120) also includes a header (122), which may contain information including the logical block number associated with the mapping block, power loss recovery (PLR) state information, or other meta-information.
Overflow blocks (130) are very similar to mapping blocks, and include a header (132) and overflow table records (134). Because the mapping blocks can hold a limited number of locators before being reclaimed, and some sectors are written much more frequently than others, the overflow blocks allow an overflow through the use of the link in the mapping table and subsequent overflow records, thus reducing reclaims of mapping blocks. Thus, the link chain for frequently written sectors will overflow into an overflow block rather than forcing an immediate reclaim of a mapping block. Overflow records may have a different number of locators per link than the mapping table records. The number of locators per link in the overflow table may be tuned to make the overflow table more efficient.
There may be two overflow regions: an active region (136) and an inactive region (138), each composed of one or more blocks. When the active overflow region is full, the active region and the inactive region are swapped, allowing a reclaim of mapping blocks. When the mapping block reclaim is completed, the now-inactive overflow blocks may be erased because there will be no links in the mapping table which point to that region.
Spare blocks (140) are empty blocks which may be used as mapping blocks in the future. Each spare block may contain a signature to identify if it has been completely erased. If the signature is not present, the spare block may be erased again.
When a logical sector number is provided, the sector number is looked up in the mapping table (202). A populated link (208) for the sector number indicates that the terminal locator for that sector is not located in the mapping table and points to the next link in the link chain. This increases efficiency, because there is no need to examine the locators under a populated link.
Here, the populated link (208) points to a record (214) in one of the overflow tables (212). The link in the overflow table is unpopulated (216). This indicates that the current record is the terminal link record, and thus, the terminal locator can be found within this record. The locator list (218) can then be scanned to determine the value of the terminal locator within the list. The terminal locator value corresponds to the physical address of the logical sector, which is the address of a slot within a data block.
When a new physical slot is allocated, the sector locator must be written to the locator link chain. A new locator and link, if necessary, are written to the locator link chain (510). The physical slot is also marked as “in use” (512). This may be accomplished by updating the block's header. Finally, the slot is programmed (514). The slot should be marked as in use after the sector locator is written so that a power loss between those events will not cause a physical slot to avoid deprecation even if a slot with the same logical number is written to the device.
Because of the write restrictions imposed by flash memory, slots and locators cannot be reprogrammed in place. Thus, reclaims must be done for both data slots and mapping blocks.
When a given physical block in memory contains a predetermined number or percentage of dirty physical slots, a data reclaim may begin. To perform a data reclaim, a spare block is taken from the spare block pool. Each valid physical slot is copied over from the original block. When all valid physical slots have been copied, the two blocks are swapped and the old block is erased in the background.
A set of physical slots is maintained as a reserve. If the amount of unallocated physical slots falls below a certain point, a background reclaim is triggered. If tuned properly, this prevents reclaims from becoming foreground operations and will help maintain consistent throughput in the system.
A mapping reclaim begins when the active overflow becomes full.
Each block in the mapping region is reclaimed one at a time. If there are blocks remaining to be reclaimed (603), a spare block is allocated from the spare block pool to replace the mapping block that is being reclaimed (604). Each index or record within the mapping block is scanned for the terminal locator (606). The terminal locator is written as the first entry in a corresponding record in the newly allocated spare block (608). After all terminal locators have been copied to the spare block, the mapping block and spare block are swapped (610) and the old mapping block is erased in the background (612).
After all of the mapping blocks are reclaimed, the newly inactive overflow blocks may be erased one at a time (614). This may be accomplished in place or by swapping an overflow block with a block from the spare block pool. Finally, the inactive region is marked as ready for use for the next reclaim cycle (616).
If the system loses power during the reclaim cycle and is reinitialized, each block's header is examined to determine its state, and reclaims are continued. This may include erasing blocks which were previously put in the spare block pool but had not yet been completely erased.
If a data reclaim is in progress when they device is powered down, the data reclaim is detected by examining the power loss recovery (PLR) status. The data reclaim may be resumed accordingly.
An in-progress mapping reclaim is detected in a similar manner. If a mapping block was being reclaimed, the entire region reclaim may be resumed rather than just reclaim of the block.
System memory or dynamic random access memory (DRAM) (706) may be coupled to the bus (710). The DRAM (706) may store an operating system (OS) (708) after system initialization.
A variety of input/output (I/O) devices (716) may be coupled to the bus (710). The I/O devices may include items such as a display, keyboard, mouse, touch screen, or other I/O devices. A wireless interface (712) may also be coupled to the bus (710). The wireless interface (712) may enable cellular or other wireless communication between the system and other devices. In one embodiment, the wireless interface (712) may include a dipole antenna.
The system also includes a flash memory device (720). The flash memory device may be built into the system, or may be part of a removable storage medium, such as a card form factor, that may be inserted into an optional flash card interface (714).
The flash memory device (720) includes a controller (722) coupled by a bus (724) to the flash array (726) and a small random access memory (RAM) (728) (e.g. several hundred bytes to a few kilobytes). Logical blocks, including data blocks, mapping blocks, overflow blocks, and spare blocks, are stored within the flash array (726) and are managed by software stored within the RAM (728).
The RAM (728) may also be used to store a small table to track physical offsets as well as information about the number of deprecated slots per block and the location of the next free slot per block.
The methods set forth above may be implemented via instructions stored on a machine-accessible medium which are executed by a processor. The instructions may be implemented in many different ways, utilizing any programming code stored on any machine-accessible medium. A machine-accessible medium includes any mechanism that provides (i.e., stores and/or transmits) information in a form readable by a machine, such as a computer. For example, a machine-accessible medium includes random-access memory (RAM), such as static RAM (SRAM) or dynamic RAM (DRAM); ROM; magnetic or optical storage medium; flash memory devices; electrical, optical, acoustical or other form of propagated signals (e.g., carrier waves, infrared signals, digital signals); etc.
Thus, sector based allocation for a flash device is disclosed. In the above description, numerous specific details are set forth. However, it is understood that embodiments may be practiced without these specific details. In other instances, well-known circuits, structures, and techniques have not been shown in detail in order not to obscure the understanding of this description. Furthermore, although some embodiments have been described as a series of operations which occur in a particular order, it may be possible to perform these operations in a different order as well. Embodiments have been described with reference to specific exemplary embodiments thereof. It will, however, be evident to persons having the benefit of this disclosure that various modifications and changes may be made to these embodiments without departing from the broader spirit and scope of the embodiments described herein. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.
