U.S. patent application number 11/342168 was filed with the patent office on 2006-08-17 for direct data file storage implementation techniques in flash memories.
Invention is credited to Alan W. Sinclair.
Application Number | 20060184722 11/342168 |
Document ID | / |
Family ID | 36694483 |
Filed Date | 2006-08-17 |
United States Patent
Application |
20060184722 |
Kind Code |
A1 |
Sinclair; Alan W. |
August 17, 2006 |
Direct data file storage implementation techniques in flash
memories
Abstract
Host system data files are written directly to a large erase
block flash memory system with a unique identification of each file
and offsets of data within the file but without the use of any
intermediate logical addresses or a virtual address space for the
memory. Directory information of where the files are stored in the
memory is maintained within the memory system by its controller,
rather than by the host. The file based interface between the host
and memory systems allows the memory system controller to utilize
the data storage blocks within the memory with increased
efficiency.
Inventors: |
Sinclair; Alan W.; (Falkirk,
GB) |
Correspondence
Address: |
PARSONS HSUE & DE RUNTZ LLP
595 MARKET STREET
SUITE 1900
SAN FRANCISCO
CA
94105
US
|
Family ID: |
36694483 |
Appl. No.: |
11/342168 |
Filed: |
January 26, 2006 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
11060248 |
Feb 16, 2005 |
|
|
|
11342168 |
Jan 26, 2006 |
|
|
|
Current U.S.
Class: |
711/103 ;
711/E12.008 |
Current CPC
Class: |
G06F 12/0246 20130101;
G06F 2212/7201 20130101; G06F 3/0607 20130101; G06F 2212/7205
20130101; G06F 3/0643 20130101; G06F 3/0679 20130101 |
Class at
Publication: |
711/103 |
International
Class: |
G06F 12/00 20060101
G06F012/00 |
Claims
1. A mass storage system, comprising: an array of non-volatile
charge storage semiconductor memory cells organized into blocks
that are erased prior to any new data being written therein and
which individually have a capacity of storing multiple units of
received data that individually include at least 512 bytes of data,
a controller connected with the memory cell array and adapted to
receive data with a logical address including a unique file
identification and offsets of data within the identified file to
program the received data at physical addresses within the blocks
of memory cells, and which further operates to maintain a plurality
of records of the programmed data for the individual files that
identify groups of variable amounts of data making up the file,
data within the groups individually having both contiguous logical
offset addresses and contiguous physical addresses, wherein the
individual records include at least a beginning logical offset
address and a beginning physical address of the data within the
group.
2. The mass storage system of claim 1, wherein the individual
records additionally includes length of the data within the
group.
3. The method of claim 1, wherein the memory cell array includes a
number of memory cells sufficient to store at least 256 megabytes
of data.
4. The method of claim 1, wherein the individual unit of data
contains at least 1024 bytes of data.
5. A mass storage system, comprising: an array of non-volatile
charge storage semiconductor memory cells organized into blocks
that are erased prior to any new data being written therein and
which individually have a capacity of storing at least 1024 bytes
of received data, a controller connected with the memory cell array
and adapted to receive data with a logical address including a
unique file identification and offsets of data within the
identified file to program the received data at physical addresses
within the blocks of memory cells as defined groups of data
individually having a size equal to or less that the capacity of
the individual memory cell blocks and containing data having
continuous logical and physical addresses, and a table capable of
storing records of individual groups of data stored in the memory
cell array, individual data records for a file of data including a
starting logical offset address, a starting physical address and
length of the data group.
6. The mass storage system of claim 5, wherein the starting
physical address of the individual data records includes an
identification of a block of memory cells and a byte location
within the identified block.
Description
CROSS REFERENCE TO RELATED APPLICATION
[0001] This application is a divisional of application Ser. No.
11/060,248, filed Feb. 16, 2005, which application is incorporated
herein in its entirety by this reference.
GENERAL BACKGROUND
[0002] This application relates to the operation of re-programmable
non-volatile memory systems such as semiconductor flash memory,
and, more specifically, to the management of the interface between
a host device and the memory. All patents, patent applications,
articles and other publications, documents and things referenced
herein are hereby incorporated herein by this reference in their
entirety for all purposes.
[0003] In an early generation of commercial flash memory systems, a
rectangular array of memory cells were divided into a large number
of groups of cells that each stored the amount of data of a
standard disk drive sector, namely 512 bytes. An additional amount
of data, such as 16 bytes, are also usually included in each group
to store an error correction code (ECC) and possibly other overhead
data relating to the user data and/or to the memory cell group in
which it is stored. The memory cells in each such group are the
minimum number of memory cells that are erasable together. That is,
the erase unit is effectively the number of memory cells that store
one data sector and any overhead data that is included. Examples of
this type of memory system are described in U.S. Pat. Nos.
5,602,987 and 6,426,893. It is a characteristic of flash memory
that the memory cells need to be erased prior to re-programming
them with data.
[0004] Flash memory systems are most commonly provided in the form
of a memory card or flash drive that is removably connected with a
variety of hosts such as a personal computer, a camera or the like,
but may also be embedded within such host systems. When writing
data to the memory, the host typically assigns unique logical
addresses to sectors, clusters or other units of data within a
continuous virtual address space of the memory system. Like a disk
operating system (DOS), the host writes data to, and reads data
from, addresses within the logical address space of the memory
system. A controller within the memory system translates logical
addresses received from the host into physical addresses within the
memory array, where the data are actually stored, and then keeps
track of these address translations. The data storage capacity of
the memory system is at least as large as the amount of data that
is addressable over the entire logical address space defined for
the memory system.
[0005] In later generations of flash memory systems, the size of
the erase unit was increased to a block of enough memory cells to
store multiple sectors of data. Even though host systems with which
the memory systems are connected may program and read data in small
minimum units such as sectors, a large number of sectors are stored
in a single erase unit of the flash memory. It is common for some
sectors of data within a block to become obsolete as the host
updates or replaces logical sectors of data. Since the entire block
must be erased before any data stored in the block can be
overwritten, new or updated data are typically stored in another
block that has been erased and has remaining capacity for the data.
This process leaves the original block with obsolete data that take
valuable space within the memory. But that block cannot be erased
if there are any valid data remaining in it.
[0006] Therefore, in order to better utilize the memory's storage
capacity, it is common to consolidate or collect valid partial
block amounts of data by copying them into an erased block so that
the block(s) from which these data are copied may then be erased
and their entire storage capacity reused. It is also desirable to
copy the data in order to group data sectors within a block in the
order of their logical addresses since this increases the speed of
reading the data and transferring the read data to the host. If
such data copying occurs too frequently, the operating performance
of the memory system can be degraded. This particularly affects
operation of memory systems where the storage capacity of the
memory is little more than the amount of data addressable by the
host through the logical address space of the system, a typical
case. In this case, data consolidation or collection may be
required before a host programming command can be executed. The
programming time is then increased.
[0007] The sizes of the blocks are increasing in successive
generations of memory systems in order to increase the number of
bits of data that may be stored in a given semiconductor area.
Blocks storing 256 data sectors and more are becoming common.
Additionally, two, four or more blocks of different arrays or
sub-arrays are often logically linked together into metablocks in
order to increase the degree of parallelism in data programming and
reading. Along with such large capacity operating units come
challenges in operating them efficiently.
SUMMARY OF THE INVENTION
[0008] Many techniques have been developed that overcome to various
degrees certain of the problems encountered in efficiently
operating such large erase block flash memory systems. The present
invention, on the other hand, takes a more fundamental approach by
changing the data transfer interface between the memory and host
system. Rather than communicating data between them by the use of
logical addresses within a virtual address space, as is currently
done, a data file is identified by a filename assigned by the host
and an offset address within the file. The memory system then knows
the host file to which each sector or other unit of data belongs.
The file unit being discussed herein is a set of data that is
ordered, such as by having sequential offset addresses, and which
is created and uniquely identified by an application program
operating in a host computing system.
[0009] This is not employed by current commercial memory systems
since hosts now identify data to the memory system within all files
by a common set of logical addresses without identifying the files.
By identifying host data by file objects instead of using logical
addresses, the memory system controller can store the data in a
manner that reduces the need for such frequent data consolidation
and collection. The frequency of data copy operations and the
amount of data copied are thus significantly reduced, thereby
increasing the data programming and reading performance of the
memory system. Further, the memory system controller maintains
directory and index table information of the memory blocks into
which host files are stored. It is then unnecessary for the host to
maintain the file allocation table (FAT) that is currently
necessary for managing a logical address interface.
[0010] Several other techniques may be employed to implement the
file-based system of the present invention, which contribute to
improved operation when used alone or together. Data from the host
are written into the memory blocks in variable length data groups
in the order the data are provided by the host. Data within each
data group have contiguous logical offset and physical memory
addresses within a block. That is, whenever a discontinuity in a
logical file offset address is encountered, a new data group is
begun. Similarly, when the data fills a memory block and begins to
be written to another, a new data group is also begun. The memory
system controller maintains directory and index table information
of the memory blocks into which the host files are stored, not the
host as in present systems. A set of entries in an index table
lists the data groups that form a complete file.
[0011] Garbage collection is initiated on a file basis, rather than
on a memory block basis as is now done, and is then performed on
all memory blocks storing data of a given file that need such an
operation. Most garbage collection can be performed in the
background, significantly reducing instances of garbage collection
having to be performed before a host write command can be
executed.
[0012] Data from an individual file are written to one or more
memory blocks dedicated to that file. The data may be written in
the order received from the host without regard to the order of
their offset addresses. The data may also be written with a
physical memory resolution that is the same as the logical offset
resolution of the file data generated by the host. One or more
common memory blocks may be maintained to store residual data from
two or more files that are left over after an integral number of
blocks have been filled with data from each file.
[0013] Other aspects, advantages, features and details of the
present invention are included in a description of exemplary
examples thereof that follows, which description should be taken in
conjunction with the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0014] FIG. 1 schematically illustrates a host and a connected
non-volatile memory system as currently implemented;
[0015] FIG. 2 is a block diagram of an example flash memory system
for use as the non-volatile memory of FIG. 1;
[0016] FIG. 3 is a representative circuit diagram of a memory cell
array that may be used in the system of FIG. 2;
[0017] FIG. 4 illustrates an example physical memory organization
of the system of FIG. 2;
[0018] FIG. 5 shows an expanded view of a portion of the physical
memory of FIG. 4;
[0019] FIG. 6 shows a further expanded view of a portion of the
physical memory of FIGS. 4 and 5;
[0020] FIG. 7 illustrates a common prior art logical address
interface between a host and a re-programmable memory system;
[0021] FIG. 8 illustrates in a different manner than FIG. 7 a
common prior art logical address interface between a host and a
re-programmable memory system;
[0022] FIG. 9 illustrates a direct file storage interface between a
host and a re-programmable memory system, according to the present
invention;
[0023] FIG. 10 illustrates in a different manner than FIG. 9 a
direct file storage interface between a host and a re-programmable
memory system, according to the present invention;
[0024] FIG. 11 shows a functional hierarchy of an example memory
system;
[0025] FIGS. 12A-12E give an example set of direct file interface
commands;
[0026] FIGS. 13A-13D show four different examples of writing data
files directly into the memory;
[0027] FIGS. 14A-14E illustrate a sequence of writing a single data
file directly into the memory;
[0028] FIG. 15 shows the result of garbage collecting the data file
illustrated in FIG. 14E;
[0029] FIG. 16 gives an example of a common block;
[0030] FIG. 17 illustrates programming a common data group into one
of several open common blocks;
[0031] FIG. 18 shows several examples of metapages of data
programmed into the non-volatile memory in different files;
[0032] FIG. 19 illustrates a structure for a file index table (FIT)
and entries from the examples of FIGS. 14A, 14C, 14E and 15;
[0033] FIG. 20 conceptually illustrates an example file indexing
data structure;
[0034] FIG. 21 shows a structure for pages of the file directory of
FIG. 20;
[0035] FIG. 22 shows a structure for pages of the file index table
of FIG. 20;
[0036] FIG. 23 shows a structure for pages of the file directory of
FIG. 20, as an alternative to that of FIG. 21;
[0037] FIG. 24 illustrates an operation of the file directories of
FIGS. 21 and 23;
[0038] FIG. 25 illustrates an operation of the file index table of
FIG. 22;
[0039] FIG. 26 is a flowchart showing an overall sequence of
operations of a memory system described herein;
[0040] FIG. 27 is a flowchart of the "Read file data" block of FIG.
26;
[0041] FIG. 28 is a flowchart of the "Program file data" block of
FIG. 26;
[0042] FIG. 29 illustrates a relative timing of two operations
included in the flowchart of FIG. 28;
[0043] FIG. 30 is a flowchart of the "Delete file" block of FIG.
26;
[0044] FIG. 31 is a flowchart of the "Garbage collection" block of
FIG. 26;
[0045] FIG. 32 is a flowchart of the "Common block garbage
collection" block of FIG. 31; and
[0046] FIG. 33 is a state diagram of the memory cell blocks during
the described operation of the example memory system herein.
FLASH MEMORY SYSTEM GENERAL DESCRIPTION
[0047] A current flash memory system and a typical operation with
host devices are described with respect to FIGS. 1-8. It is in such
a system that the various aspects of the present invention may be
implemented. A host system 1 of FIG. 1 stores data into and
retrieves data from a flash memory 2. Although the flash memory can
be embedded within the host, the memory 2 is illustrated to be in
the more popular form of a card that is removably connected to the
host through mating parts 3 and 4 of a mechanical and electrical
connector. There are currently many different flash memory cards
that are commercially available, examples being the CompactFlash
(CF), the MultiMediaCard (MMC), Secure Digital (SD), miniSD, Memory
Stick, SmartMedia and TransFlash cards. Although each of these
cards has a unique mechanical and/or electrical interface according
to its standardized specifications, the flash memory included in
each is very similar. These cards are all available from SanDisk
Corporation, assignee of the present application. SanDisk also
provides a line of flash drives under its Cruzer trademark, which
are hand held memory systems in small packages that have a
Universal Serial Bus (USB) plug for connecting with a host by
plugging into the host's USB receptacle. Each of these memory cards
and flash drives includes controllers that interface with the host
and control operation of the flash memory within them.
[0048] Host systems that use such memory cards and flash drives are
many and varied. They include personal computers (PCs), laptop and
other portable computers, cellular telephones, personal digital
assistants (PDAs), digital still cameras, digital movie cameras and
portable audio players. The host typically includes a built-in
receptacle for one or more types of memory cards or flash drives
but some require adapters into which a memory card is plugged.
[0049] The host system 1 of FIG. 1 may be viewed as having two
major parts, insofar as the memory 2 is concerned, made up of a
combination of circuitry and software. They are an applications
portion 5 and a driver portion 6 that interfaces with the memory 2.
In a personal computer, for example, the applications portion 5 can
include a processor running word processing, graphics, control or
other popular application software. In a camera, cellular telephone
or other host system that is primarily dedicated to performing a
single set of functions, the applications portion 5 includes the
software that operates the camera to take and store pictures, the
cellular telephone to make and receive calls, and the like.
[0050] The memory system 2 of FIG. 1 includes flash memory 7, and
circuits 8 that both interface with the host to which the card is
connected for passing data back and forth and control the memory 7.
The controller 8 typically converts between logical addresses of
data used by the host 1 and physical addresses of the memory 7
during data programming and reading.
[0051] Referring to FIG. 2, circuitry of a typical flash memory
system that may be used as the non-volatile memory 2 of FIG. 1 is
described. The system controller is usually implemented on a single
integrated circuit chip 11 that is connected in parallel with one
or more integrated circuit memory chips over a system bus 13, a
single such memory chip 15 being shown in FIG. 2. The particular
bus 13 that is illustrated includes a separate set of conductors 17
to carry data, a set 19 for memory addresses and a set 21 for
control and status signals. Alternatively, a single set of
conductors may be time shared between these three functions.
Further, other configurations of system buses can be employed, such
as a ring bus that is described in U.S. patent application Ser. No.
10/915,039, filed Aug. 9, 2004, entitled "Ring Bus Structure and
It's Use in Flash Memory Systems."
[0052] A typical controller chip 11 has its own internal bus 23
that interfaces with the system bus 13 through interface circuits
25. The primary functions normally connected to the bus are a
processor 27 (such as a microprocessor or micro-controller), a
read-only-memory (ROM) 29 containing code to initialize ("boot")
the system, read-only-memory (RAM) 31 used primarily to buffer data
being transferred between the memory and a host, and circuits 33
that calculate and check an error correction code (ECC) for data
passing through the controller between the memory and the host. The
controller bus 23 interfaces with a host system through circuits
35, which, in the case of the system of FIG. 2 being contained
within a memory card, is done through external contacts 37 of the
card that are part of the connector 4. A clock 39 is connected with
and utilized by each of the other components of the controller
11.
[0053] The memory chip 15, as well as any other connected with the
system bus 13, typically contains an array of memory cells
organized into multiple sub-arrays or planes, two such planes 41
and 43 being illustrated for simplicity but more, such as four or
eight such planes, may instead be used. Alternatively, the memory
cell array of the chip 15 may not be divided into planes. When so
divided however, each plane has its own column control circuits 45
and 47 that are operable independently of each other. The circuits
45 and 47 receive addresses of their respective memory cell array
from the address portion 19 of the system bus 13, and decode them
to address a specific one or more of respective bit lines 49 and
51. The word lines 53 are addressed through row control circuits 55
in response to addresses received on the address bus 19. Source
voltage control circuits 57 and 59 are also connected with the
respective planes, as are p-well voltage control circuits 61 and
63. If the memory chip 15 has a single array of memory cells, and
if two or more such chips exist in the system, the array of each
chip may be operated similarly to a plane or sub-array within the
multi-plane chip described above.
[0054] Data are transferred into and out of the planes 41 and 43
through respective data input/output circuits 65 and 67 that are
connected with the data portion 17 of the system bus 13. The
circuits 65 and 67 provide for both programming data into the
memory cells and for reading data from the memory cells of their
respective planes, through lines 69 and 71 connected to the planes
through respective column control circuits 45 and 47.
[0055] Although the controller 11 controls the operation of the
memory chip 15 to program data, read data, erase and attend to
various housekeeping matters, each memory chip also contains some
controlling circuitry that executes commands from the controller 11
to perform such functions. Interface circuits 73 are connected to
the control and status portion 21 of the system bus 13. Commands
from the controller are provided to a state machine 75 that then
provides specific control of other circuits in order to execute
these commands. Control lines 77-81 connect the state machine 75
with these other circuits as shown in FIG. 2. Status information
from the state machine 75 is communicated over lines 83 to the
interface 73 for transmission to the controller 11 over the bus
portion 21.
[0056] A NAND architecture of the memory cell arrays 41 and 43 is
currently preferred, although other architectures, such as NOR, can
also be used instead. Examples of NAND flash memories and their
operation as part of a memory system may be had by reference to
U.S. Pat. Nos. 5,570,315, 5,774,397, 6,046,935, 6,373,746,
6,456,528, 6,522,580, 6,771,536 and 6,781,877 and United States
patent application publication no. 2003/0147278.
[0057] An example NAND array is illustrated by the circuit diagram
of FIG. 3, which is a portion of the memory cell array 41 of the
memory system of FIG. 2. A large number of global bit lines are
provided, only four such lines 91-94 being shown in FIG. 2 for
simplicity of explanation. A number of series connected memory cell
strings 97-104 are connected between one of these bit lines and a
reference potential. Using the memory cell string 99 as
representative, a plurality of charge storage memory cells 107-110
are connected in series with select transistors 111 and 112 at
either end of the string. When the select transistors of a string
are rendered conductive, the string is connected between its bit
line and the reference potential. One memory cell within that
string is then programmed or read at a time.
[0058] Word lines 115-118 of FIG. 3 individually extend across the
charge storage element of one memory cell in each of a number of
strings of memory cells, and gates 119 and 120 control the states
of the select transistors at each end of the strings. The memory
cell strings that share common word and control gate lines 115-120
are made to form a block 123 of memory cells that are erased
together. This block of cells contains the minimum number of cells
that are physically erasable at one time. One row of memory cells,
those along one of the word lines 115-118, are programmed at a
time. Typically, the rows of a NAND array are programmed in a
prescribed order, in this case beginning with the row along the
word line 118 closest to the end of the strings connected to ground
or another common potential. The row of memory cells along the word
line 117 is programmed next, and so on, throughout the block 123.
The row along the word line 115 is programmed last.
[0059] A second block 125 is similar, its strings of memory cells
being connected to the same global bit lines as the strings in the
first block 123 but having a different set of word and control gate
lines. The word and control gate lines are driven to their proper
operating voltages by the row control circuits 55. If there is more
than one plane or sub-array in the system, such as planes 1 and 2
of FIG. 2, one memory architecture uses common word lines extending
between them. There can alternatively be more than two planes or
sub-arrays that share common word lines. In other memory
architectures, the word lines of individual planes or sub-arrays
are separately driven.
[0060] As described in several of the NAND patents and published
application referenced above, the memory system may be operated to
store more than two detectable levels of charge in each charge
storage element or region, thereby to store more than one bit of
data in each. The charge storage elements of the memory cells are
most commonly conductive floating gates but may alternatively be
non-conductive dielectric charge trapping material, as described in
United States patent application publication no. 2003/0109093.
[0061] FIG. 4 conceptually illustrates an organization of the flash
memory cell array 7 (FIG. 1) that is used as an example in further
descriptions below. Four planes or sub-arrays 131-134 of memory
cells may be on a single integrated memory cell chip, on two chips
(two of the planes on each chip) or on four separate chips. The
specific arrangement is not important to the discussion below. Of
course, other numbers of planes, such as 1, 2, 8, 16 or more may
exist in a system. The planes are individually divided into blocks
of memory cells shown in FIG. 4 by rectangles, such as blocks 137,
138, 139 and 140, located in respective planes 131-134. There can
be dozens or hundreds of blocks in each plane. As mentioned above,
the block of memory cells is the unit of erase, the smallest number
of memory cells that are physically erasable together. For
increased parallelism, however, the blocks are operated in larger
metablock units. One block from each plane is logically linked
together to form a metablock. The four blocks 137-140 are shown to
form one metablock 141. All of the cells within a metablock are
typically erased together. The blocks used to form a metablock need
not be restricted to the same relative locations within their
respective planes, as is shown in a second metablock 143 made up of
blocks 145-148. Although it is usually preferable to extend the
metablocks across all of the planes, for high system performance,
the memory system can be operated with the ability to dynamically
form metablocks of any or all of one, two or three blocks in
different planes. This allows the size of the metablock to be more
closely matched with the amount of data available for storage in
one programming operation.
[0062] The individual blocks are in turn divided for operational
purposes into pages of memory cells, as illustrated in FIG. 5. The
memory cells of each of the blocks 131-134, for example, are each
divided into eight pages P0-P7. Alternatively, there may be 16, 32
or more pages of memory cells within each block. The page is the
unit of data programming and reading within a block, containing the
minimum amount of data that are programmed at one time. In the NAND
architecture of FIG. 3, a page is formed of memory cells along a
word line within a block. However, in order to increase the memory
system operational parallelism, such pages within two or more
blocks may be logically linked into metapages. A metapage 151 is
illustrated in FIG. 5, being formed of one physical page from each
of the four blocks 131-134. The metapage 151, for example, includes
the page P2 in of each of the four blocks but the pages of a
metapage need not necessarily have the same relative position
within each of the blocks. Although it is preferable to program and
read the maximum amount of data in parallel across all four planes,
for high system performance, the memory system can also be operated
to form metapages of any or all of one, two or three pages in
separate blocks in different planes. This allows the programming
and reading operations to adaptively match the amount of data that
may be conveniently handled in parallel and reduces the occasions
when part of a metapage remains unprogrammed with data.
[0063] A metapage formed of physical pages of multiple planes, as
illustrated in FIG. 5, contains memory cells along word line rows
of those multiple planes. Rather than programming all of the cells
in one word line row at the same time, they are more commonly
alternately programmed in two or more interleaved groups, each
group storing a page of data (in a single block) or a metapage of
data (across multiple blocks). By programming alternate memory
cells at one time, a unit of peripheral circuits including data
registers and a sense amplifier need not be provided for each bit
line but rather are time-shared between adjacent bit lines. This
economizes on the amount of substrate space required for the
peripheral circuits and allows the memory cells to be packed with
an increased density along the rows. Otherwise, it is preferable to
simultaneously program every cell along a row in order to maximize
the parallelism available from a given memory system.
[0064] With reference to FIG. 3, the simultaneous programming of
data into every other memory cell along a row is most conveniently
accomplished by providing two rows of select transistors (not
shown) along at least one end of the NAND strings, instead of the
single row that is shown. The select transistors of one row then
connect every other string within a block to their respective bit
lines in response to one control signal, and the select transistors
of the other row connect intervening every other string to their
respective bit lines in response to another control signal. Two
pages of data are therefore written into each row of memory
cells.
[0065] The amount of data in each logical page is typically an
integer number of one or more sectors of data, each sector
containing 512 bytes of data, by convention. FIG. 6 shows a logical
data page of two sectors 153 and 155 of data of a page or metapage.
Each sector usually contains a portion 157 of 512 bytes of user or
system data being stored and another number of bytes 159 for
overhead data related either to the data in the portion 157 or to
the physical page or block in which it is stored. The number of
bytes of overhead data is typically 16 bytes, making the total 528
bytes for each of the sectors 153 and 155. The overhead portion 159
may contain an ECC calculated from the data portion 157 during
programming, its logical address, an experience count of the number
of times the block has been erased and re-programmed, one or more
control flags, operating voltage levels, and/or the like, plus an
ECC calculated from such overhead data 159. Alternatively, the
overhead data 159, or a portion of it, may be stored in different
pages in other blocks.
[0066] As the parallelism of memories increases, data storage
capacity of the metablock increases and the size of the data page
and metapage also increase as a result. The data page may then
contain more than two sectors of data. With two sectors in a data
page, and two data pages per metapage, there are four sectors in a
metapage. Each metapage thus stores 2048 bytes of data. This is a
high degree of parallelism, and can be increased even further as
the number of memory cells in the rows are increased. For this
reason, the width of flash memories is being extended in order to
increase the amount of data in a page and a metapage.
[0067] The physically small re-programmable non-volatile memory
cards and flash drives identified above are commercially available
with data storage capacity of 512 megabytes (MB), 1 gigabyte (GB),
2 GB and 4 GB, and may go higher. FIG. 7 illustrates the most
common interface between a host and such a mass memory system. The
host deals with data files generated or used by application
software or firmware programs executed by the host. A word
processing data file is an example, and a drawing file of computer
aided design (CAD) software is another, found mainly in general
computer hosts such as PCs, laptop computers and the like. A
document in the pdf format is also such a file. A still digital
video camera generates a data file for each picture that is stored
on a memory card. A cellular telephone utilizes data from files on
an internal memory card, such as a telephone directory. A PDA
stores and uses several different files, such as an address file, a
calendar file, and the like. In any such application, the memory
card may also contain software that operates the host.
[0068] A common logical interface between the host and the memory
system is illustrated in FIG. 7. A continuous logical address space
161 is large enough to provide addresses for all the data that may
be stored in the memory system. The host address space is typically
divided into increments of clusters of data. Each cluster may be
designed in a given host system to contain a number of sectors of
data, somewhere between 4 and 64 sectors being typical. A standard
sector contains 512 bytes of data.
[0069] Three Files 1, 2 and 3 are shown in the example of FIG. 7 to
have been created. An application program running on the host
system creates each file as an ordered set of data and identifies
it by a unique name or other reference. Enough available logical
address space not already allocated to other files is assigned by
the host to File 1. File 1 is shown to have been assigned a
contiguous range of available logical addresses. Ranges of
addresses are also commonly allocated for specific purposes, such
as a particular range for the host operating software, which are
then avoided for storing data even if these addresses have not been
utilized at the time the host is assigning logical addresses to the
data.
[0070] When a File 2 is later created by the host, the host
similarly assigns two different ranges of contiguous addresses
within the logical address space 161, as shown in FIG. 7. A file
need not be assigned contiguous logical addresses but rather can be
fragments of addresses in between address ranges already allocated
to other files. This example then shows that yet another File 3
created by the host is allocated other portions of the host address
space not previously allocated to the Files 1 and 2 and other
data.
[0071] The host keeps track of the memory logical address space by
maintaining a file allocation table (FAT), where the logical
addresses the host assigns to the various host files are
maintained. The FAT table is typically stored in the non-volatile
memory, as well as in a host memory, and is frequently updated by
the host as new files are stored, other files deleted, files
modified and the like. When a host file is deleted, for example,
the host then deallocates the logical addresses previously
allocated to the deleted file by updating the FAT table to show
that they are now available for use with other data files.
[0072] The host is not concerned about the physical locations where
the memory system controller chooses to store the files. The
typical host only knows its logical address space and the logical
addresses that it has allocated to its various files. The memory
system, on the other hand, through a typical host/card interface,
only knows the portions of the logical address space to which data
have been written but does not know the logical addresses allocated
to specific host files, or even the number of host files. The
memory system controller converts the logical addresses provided by
the host for the storage or retrieval of data into unique physical
addresses within the flash memory cell array where host data are
stored. A block 163 represents a working table of these
logical-to-physical address conversions, which is maintained by the
memory system controller.
[0073] The memory system controller is programmed to store data
files within the blocks and metablocks of a memory array 165 in a
manner to maintain the performance of the system at a high level.
Four planes or sub-arrays are used in this illustration. Data are
preferably programmed and read with the maximum degree of
parallelism that the system allows, across an entire metablock
formed of a block from each of the planes. At least one metablock
167 is usually allocated as a reserved block for storing operating
firmware and data used by the memory controller. Another metablock
169, or multiple metablocks, may be allocated for storage of host
operating software, the host FAT table and the like. Most of the
physical storage space remains for the storage of data files. The
memory controller does not know, however, how the data received has
been allocated by the host among its various file objects. All the
memory controller typically knows from interacting with the host is
that data written by the host to specific logical addresses are
stored in corresponding physical addresses as maintained by the
controller's logical-to-physical address table 163.
[0074] In a typical memory system, a few extra blocks of storage
capacity are provided than are necessary to store the amount of
data within the address space 161. One or more of these extra
blocks may be provided as redundant blocks for substitution for
other blocks that may become defective during the lifetime of the
memory. The logical grouping of blocks contained within individual
metablocks may usually be changed for various reasons, including
the substitution of a redundant block for a defective block
originally assigned to the metablock. One or more additional
blocks, such as metablock 171, are typically maintained in an
erased block pool. When the host writes data to the memory system,
the controller converts the logical addresses assigned by the host
to physical addresses within a metablock in the erased block pool.
Other metablocks not being used to store data within the logical
address space 161 are then erased and designated as erased pool
blocks for use during a subsequent data write operation.
[0075] Data stored at specific host logical addresses are
frequently overwritten by new data as the original stored data
become obsolete. The memory system controller, in response, writes
the new data in an erased block and then changes the
logical-to-physical address table for those logical addresses to
identify the new physical block to which the data at those logical
addresses are stored. The blocks containing the original data at
those logical addresses are then erased and made available for the
storage of new data. Such erasure often must take place before a
current data write operation may be completed if there is not
enough storage capacity in the pre-erased blocks from the erase
block pool at the start of writing. This can adversely impact the
system data programming speed. The memory controller typically
learns that data at a given logical address has been rendered
obsolete by the host only when the host writes new data to their
same logical address. Many blocks of the memory can therefore be
storing such invalid data for a time.
[0076] The sizes of blocks and metablocks are increasing in order
to efficiently use the area of the integrated circuit memory chip.
This results in a large proportion of individual data writes
storing an amount of data that is less than the storage capacity of
a metablock, and in many cases even less than that of a block.
Since the memory system controller normally directs new data to an
erased pool metablock, this can result in portions of metablocks
going unfilled. If the new data are updates of some data stored in
another metablock, remaining valid metapages of data from that
other metablock having logical addresses contiguous with those of
the new data metapages are also desirably copied in logical address
order into the new metablock. The old metablock may retain other
valid data metapages. This results over time in data of certain
metapages of an individual metablock being rendered obsolete and
invalid, and replaced by new data with the same logical address
being written to a different metablock.
[0077] In order to maintain enough physical memory space to store
data over the entire logical address space 161, such data are
periodically compacted or consolidated (garbage collection). It is
also desirable to maintain sectors of data within the metablocks in
the same order as their logical addresses as much as practical,
since this makes reading data in contiguous logical addresses more
efficient. So data compaction and garbage collection are typically
performed with this additional goal. Some aspects of managing a
memory when receiving partial block data updates and the use of
metablocks are described in U.S. Pat. No. 6,763,424.
[0078] Data compaction typically involves reading all valid data
metapages from a metablock and writing them to a new block,
ignoring metapages with invalid data in the process. The metapages
with valid data are also preferably arranged with a physical
address order that matches the logical address order of the data
stored in them. The number of metapages occupied in the new
metablock will be less than those occupied in the old metablock
since the metapages containing invalid data are not copied to the
new metablock. The old block is then erased and made available to
store new data. The additional metapages of capacity gained by the
consolidation can then be used to store other data.
[0079] During garbage collection, metapages of valid data with
contiguous or near contiguous logical addresses are gathered from
two or more metablocks and re-written into another metablock,
usually one in the erased block pool. When all valid data metapages
are copied from the original two or more metablocks, they may be
erased for future use.
[0080] Data consolidation and garbage collection take time and can
affect the performance of the memory system, particularly if data
consolidation or garbage collection needs to take place before a
command from the host can be executed. Such operations are normally
scheduled by the memory system controller to take place in the
background as much as possible but the need to perform these
operations can cause the controller to have to give the host a busy
status signal until such an operation is completed. An example of
where execution of a host command can be delayed is where there are
not enough pre-erased metablocks in the erased block pool to store
all the data that the host wants to write into the memory and data
consolidation or garbage collection is needed first to clear one or
more metablocks of valid data, which can then be erased. Attention
has therefore been directed to managing control of the memory in
order to minimize such disruptions. Many such techniques are
described in the following U.S. patent applications: Ser. No.
10/749,831, filed Dec. 30, 2003, entitled "Management of
Non-Volatile Memory Systems Having Large Erase Blocks"; Ser. No.
10/750,155, filed Dec. 30, 2003, entitled "Non-Volatile Memory and
Method with Block Management System"; Ser. No. 10/917,888, filed
Aug. 13, 2004, entitled "Non-Volatile Memory and Method with Memory
Planes Alignment"; Ser. No. 10/917,867, filed Aug. 13, 2004; Ser.
No. 10/917,889, filed Aug. 13, 2004, entitled "Non-Volatile Memory
and Method with Phased Program Failure Handling"; and Ser. No.
10/917,725, filed Aug. 13, 2004, entitled "Non-Volatile Memory and
Method with Control Data Management."
[0081] One challenge to efficiently control operation of memory
arrays with very large erase blocks is to match and align the
number of data sectors being stored during a given write operation
with the capacity and boundaries of blocks of memory. One approach
is to configure a metablock used to store new data from the host
with less than a maximum number of blocks, as necessary to store a
quantity of data less than an amount that fills an entire
metablock. The use of adaptive metablocks is described in U.S.
patent application Ser. No. 10/749,189, filed Dec. 30, 2003,
entitled "Adaptive Metablocks." The fitting of boundaries between
blocks of data and physical boundaries between metablocks is
described in patent application Ser. No. 10/841,118, filed May 7,
2004, and Ser. No. 11/016,271, filed Dec. 16, 2004, entitled "Data
Run Programming."
[0082] The memory controller may also use data from the FAT table,
which is stored by the host in the non-volatile memory, to more
efficiently operate the memory system. One such use is to learn
when data has been identified by the host to be obsolete by
deallocating their logical addresses. Knowing this allows the
memory controller to schedule erasure of the blocks containing such
invalid data before it would normally learn of it by the host
writing new data to those logical addresses. This is described in
U.S. patent application Ser. No. 10/897,049, filed Jul. 21, 2004,
entitled "Method and Apparatus for Maintaining Data on Non-Volatile
Memory Systems." Other techniques include monitoring host patterns
of writing new data to the memory in order to deduce whether a
given write operation is a single file, or, if multiple files,
where the boundaries between the files lie. U.S. patent application
Ser. No. 11/022,369, filed Dec. 23, 2004, entitled "FAT Analysis
for Optimized Sequential Cluster Management," describes the use of
techniques of this type.
[0083] To operate the memory system efficiently, it is desirable
for the controller to know as much about the logical addresses
assigned by the host to data of its individual files as it can.
Data files can then be stored by the controller within a single
metablock or group of metablocks, rather than being scattered among
a larger number of metablocks when file boundaries are not known.
The result is that the number and complexity of data consolidation
and garbage collection operations are reduced. The performance of
the memory system improves as a result. But it is difficult for the
memory controller to know much about the host data file structure
when the host/memory interface includes the logical address space
161 (FIG. 7), as described above.
[0084] Referring to FIG. 8, the typical logical address host/memory
interface as already shown in FIG. 7 is illustrated differently.
The host generated data files are allocated logical addresses by
the host. The memory system then sees these logical addresses and
maps them into physical addresses of blocks of memory cells where
the data are actually stored.
DESCRIPTION OF EXEMPLARY FILE-BASED INTERFACE EMBODIMENTS
[0085] The improved interface between a host and memory system for
the storage of mass amounts of data eliminates use of the logical
address space. The host instead logically addresses each file by a
unique field (or other unique reference) and offset addresses of
units of data (such as bytes) within the file. This file address is
given directly to the memory system controller, which then keeps
its own table of where the data of each host file are physically
stored. This new interface can be implemented with the same memory
system as described above with respect to FIGS. 2-6. The primary
difference with what is described above is the manner in which that
memory system communicates with a host system.
[0086] This file-based interface is illustrated in FIG. 9, which
should be compared with the logical address interface of FIG. 7. An
identification of each of the Files 1, 2 and 3 and offsets of data
within the files of FIG. 9 are passed directly to the V memory
controller. This logical address information is then translated by
a memory controller function 173 into physical addresses of
metablocks and metapages of the memory 165.
[0087] The file-based interface is also illustrated by FIG. 10,
which should be compared with the logical address interface of FIG.
8. The logical address space and host maintained FAT table of FIG.
8 are not present in FIG. 10. Rather, data files generated by the
host are identified to the memory system by file number and offsets
of data within the file. The memory system then directly maps the
files to the physical blocks of the memory cell array.
[0088] With reference to FIG. 11, functional layers of an example
mass storage system being described herein are illustrated. It is
the "Direct File Storage Back End System" that is primarily the
subject of this description. This is internal to operation of the
memory system, and communicates through a "Direct-File Interface"
and a "File-Based Front-End System" with a host system over a
file-based interface channel. Each host file is uniquely
identified, such as by a file name. Data within a file are
identified by an offset address within a linear address space that
is unique to the file. There is no need for a logical address space
to be defined for the memory system.
Commands
[0089] A set of direct file interface commands from the host system
supports the operation of the memory system. An example set of such
commands is given in FIGS. 12A-12E. These are only briefly
summarized here, for reference throughout the remaining portion of
this description. FIG. 12A list the host commands used to cause
data to be transferred between the host and memory systems,
according to a defined protocol. Data within a designated file
(<fileID>) at a particular offset (<offset>) within the
file is either written to or read from the memory system.
Transmission of a Write, Insert or Update command is followed by
transmission of data from the host to the memory system, and the
memory system responds by writing the data in its memory array.
Transmission of a Read command by the host causes the memory system
to respond by sending data of the designated file to the host. A
data offset need not be sent with the Write command if the memory
system maintains a pointer identifying the next storage location
where additional data of the file may be stored. However, if a
Write command includes an offset address within the file already
written, the memory device may interpret that to be a command to
update the file data beginning at the offset address, thereby
eliminating the need for a separate Update command. For the Read
command, a data offset need not be specified by the host if the
entire file is to be read. Execution of one of these FIG. 12A data
commands is terminated in response to the transmission by the host
system of any other command.
[0090] Another data command is a Remove command. Unlike the other
data commands of FIG. 12A, the Remove command is not followed by
the transmission of data. Its effect is to cause the memory system
to mark data between the specified offset1 and offset2 as obsolete.
These data are then removed during the next data compaction or
garbage collection of the file or block in which the obsolete data
exits.
[0091] FIG. 12B lists host commands that manage files within the
memory system. When the host is about to write data of a new file
in the memory system, it first issues an Open command and the
memory system responds by opening a new file. A number of files
that can remain open at one time will usually be specified. When
the host closes a file, a Close command tells the memory system
that its resources used to maintain the open file can be
redirected. The memory system will typically immediately schedule
such a file for garbage collection. With the direct file interface
being described, garbage collection is logically managed and
performed primarily on files, not physically with individual memory
cell blocks. The Close_after command gives the memory system
advanced notice that a file is about to be closed. The file Delete
command causes the memory system to immediately schedule the memory
cell blocks containing data from the deleted file to be erased, in
accordance with specified priority rules. An Erase command
specifies that data of the specified file be immediately erased
from the memory.
[0092] The primary difference between the Delete and Erase commands
is the priority given to erasing the designated file data. The host
may use the Erase command to remove secure or sensitive data from
the memory at the earliest practical time, while the Delete command
causes such data to be erased with a lower priority. Use of the
Erase command when powering down the memory system removes
sensitive data before the memory device is removed from the host
and thus prevents dissemination of that data to other users or host
systems during a subsequent use of the memory device. Both of these
commands are preferably executed in the background; i.e., without
slowing down execution of the primary data commands (FIG. 12A). In
any event, receipt of another command from the host will usually
cause the memory controller to terminate any background
operation.
[0093] Host commands that relate to directories within the memory
system are listed in FIG. 12C. Each directory command includes an
identification (<directoryID>) of the directory to which the
command pertains. Although the memory system controller maintains
the directories, commands with respect to the directories and
designations of the directories are provided by the host system.
The memory controller executes these commands, with the host
supplied directory designations, pursuant to the firmware stored in
the memory system.
[0094] The <fileID> parameter can be either a full pathname
for the file, or some shorthand identifier for the file, referenced
herein as a file_handle. A file pathname is provided to the
Direct-File Interface of FIG. 11 in association with certain
commands. This allows a fully explicit entry to be created in the
file directory when a file is opened for the first time, and allows
the correct existing entry in the file directory to be accessed
when an existing file is opened. The file pathname syntax may
conform to the standard used by the DOS file system. The pathname
describes a hierarchy of directories and a file within the lowest
level of directory. Path segments may be delimited by "\". A path
prefixed by "\" is relative to the root directory. A path not
prefixed by "\" is relative to the current directory. A segment
path of ".." indicates the parent directory of the current
directory.
[0095] Open files may alternatively be identified by a file_handle
parameter, which is assigned by the storage device when the file is
first created. The storage device can then communicate the
shorthand file designation to the host each time the host opens the
file. The host may then use the file_handle with the Write, Insert,
Update, Read, Close and Close_after commands of an open file.
Access to the file by the host will typically be quicker than if a
full pathname is used since the hierarchy of the file directory
need not be navigated. When a file is first opened by use of the
Open command, the full pathname is usually used since a file_handle
has likely not yet been assigned to that file by the memory system.
But a file_handle can be used if already available. For the
remaining Delete and Erase commands of FIGS. 12A and 12B that
utilize a fileID, use of a complete file pathname is preferred as
security against an incorrect file_handle being supplied by the
host. It is more difficult for the host to inadvertently generate
an incorrect pathname that matches one of an existing but
unintended file.
[0096] The directory commands of FIG. 12C are similarly received by
the Direct-File Interface of FIG. 11 with a <directoryID>
identification of the directory to which they pertain. A full
pathname is the preferred directoryID that is received with a
directory command.
[0097] The file_handle is a shortform identifier that is returned
at the Direct-File Interface to the host by the mass storage device
in response to an Open command. It is convenient to define the
file_handle as being the pointer to the FIT that exists in the
directory entry for the file. This pointer defines the logical FIT
block number and logical file number within that block for the
file. Using this as a file_handle allows the file FIT entries to be
accessed without first having to search for the file in the file
directory. For example, if the memory device can have up to 64 FIT
blocks, and each FIT block can index up to 64 files, then a file
with file_handle 1107 has the pointer to its data group entries in
the FIT set to logical fife 7 in FIT block 11. This file_handle is
generated by the memory system controller when directory and FIT
entries for a file are created in response to an Open command and
becomes invalid in response to a Close command.
[0098] FIG. 12D give host commands that manage the state of the
interface between the host and memory systems. The Idle command
tells the memory system that it may perform internal operations
such as data erasure and garbage collection that have previously
been scheduled. In response to receiving the Standby command, the
memory system will stop performing background operations such as
garbage collection and data erasure. The Shut-down command gives
the memory controller advance warning of an impending loss of
power, which allows completion of pending memory operations
including writing data from volatile controller buffers into
non-volatile flash memory.
[0099] A Size command, shown in FIG. 12E, will typically be issued
by a host before a Write command. The memory system, in response,
reports to the host the available capacity for further file data to
be written. This may be calculated on the basis of available
unprogrammed physical capacity minus physical capacity required to
manage storage of the defined file data capacity.
[0100] When the host issues a Status command (FIG. 12E), the memory
device will respond with its current status. This response may be
in the form of a binary word or words with different fields of bits
providing the host with various specific items of information about
the memory device. For example, one two-bit field can report
whether the device is busy, and, if so, provide more than one busy
status depending upon what the memory device is busy doing. One
busy status can indicate that the memory device is dealing with
executing a host write or read command to transfer data, a
foreground operation. A second busy status indication can be used
to tell the host when the memory system is performing a background
housekeeping operation, such as data compaction or garbage
collection. The host can decide whether to wait until the end of
this second busy before sending another command to the memory
device. If another command is sent before the housekeeping
operation is completed, the memory device will end the housekeeping
operation and execute the command.
[0101] The host can use the second device busy in combination with
the Idle command to allow housekeeping operations to take place
within the memory device. After the host sends a command, or a
series of commands, that likely creates the need for the device to
do a housekeeping operation, the host can send the Idle command. As
described later, the memory device can be programmed to respond to
an Idle command by initiating a housekeeping operation and at the
same time start the second busy described above. A Delete command,
for example, creates the need to perform garbage collection,
according to the algorithms described below. An Idle command from
the host after having issued a series of Delete commands then
allows the device time to perform garbage collection that may be
necessary for the memory device to be able to respond to a
subsequent host Write command. Otherwise, the garbage collection
may need to be performed after receiving the next Write command but
before it can be executed, thereby significantly slowing down
execution of that command.
Writing Data
[0102] When a new data file is programmed into the memory, the data
are written into an erased block of memory cells beginning with the
first physical location in the block and proceeding through the
locations of the block sequentially in order. The data are
programmed in the order received from the host, regardless of the
order of the offsets of that data within the file. Programming
continues until all data of the file have been written into the
memory. If the amount of data in the file exceeds the capacity of a
single memory block, then, when the first block is full,
programming continues in a second erased block. The second memory
block is programmed in the same manner as the first, in order from
the first location until either all the data of the file are stored
or the second block is full. A third or additional blocks may be
programmed with any remaining data of the file. Multiple blocks or
metablocks storing data of a single file need not be physically or
logically contiguous. For ease of explanation, unless otherwise
specified, it is intended that the term "block" as used herein
refer to either the block unit of erase or a multiple block
"metablock," depending upon whether metablocks are being used in a
specific system.
[0103] Referring to FIG. 13A, the writing of a data file to the
memory system is illustrated. A data file 181, in this example, is
larger than the storage capacity of one block or metablock 183 of
the memory system, which is shown to extend between solid vertical
lines. A portion 184 of the data file 181 is therefore also written
into a second block 185. These memory cell blocks are shown to be
physically contiguous but they need not be. Data from the file 181
are written as they are received streaming from the host until all
the data of the file have been written into the memory. In the
example of FIG. 13A, the data 181 are the initial data for file,
received from the host after a Write command of FIG. 12A.
[0104] A preferred way for the memory system to manage and keep
track of the stored data is with the use of variable sized data
groups. That is, data of a file are stored as a plurality of groups
of data that may be chained together in a defined order to form the
complete file. Preferably, however, the order of the data groups
within the file is maintained by the memory system controller
through use of a file index table (FIT). As a stream of data from
the host are being written, a new data group is begun whenever
there is a discontinuity either in the logical offset addresses of
the file data or in the physical space in which the data are being
stored. An example of such a physical discontinuity is when data of
a file fills one block and begins to be written into another block.
This is illustrated in FIG. 13A, wherein a first data group fills
the first block 183 the remaining portion 184 of the file is stored
in the second block 185 as a second data group. The first data
group can be represented by (F0,D0), where F0 is the logical offset
of the beginning of the data file and D0 is the physical location
within memory where the file begins. The second data group is
represented as (F1,D1), where F1 is the logical file offset of data
that is stored at the beginning of the second block 185 and D1 is
the physical location where that data are stored.
[0105] The amount of data being transferred through the host-memory
interface may be expressed in terms of a number of bytes of data, a
number of sectors of data, or with some other granularity. A host
most often defines data of its files with byte granularity but then
groups bytes into sectors of 512 bytes each, or into clusters of
multiple sectors each, when communicating with a large capacity
memory system through a current logical address interface. This is
usually done to simplify operation of the memory system. Although
the file-based host-memory interface being described herein may use
some other unit of data, the original host file byte granularity is
generally preferred. That is, data offsets, lengths, and the like,
are preferably expressed in terms of byte(s), the smallest
reasonable unit of data, rather than by sector(s), cluster(s) or
the like. This allows more efficient use of the capacity of the
flash memory storage with the techniques described herein.
[0106] In common existing logical address interfaces, the host also
specifies the length of the data being written. This can also be
done with the file-based interface described herein but since it is
not necessary for execution of the Write command (FIG. 12A), it is
preferred that the host not provide the length of data being
written.
[0107] The new file written into the memory in the manner
illustrated in FIG. 13A is then represented in a FIT as a sequence
of index entries (F0,D0), (F1,D1) for the data groups, in that
order. That is, whenever the host system wants to access a
particular file, the host sends its fileID or other identification
to the memory system, which then accesses its FIT to identify the
data groups that make up that file. The length <length> of
the individual data groups may also be included in their individual
entries, for convenience of operation of the memory system. When
used, the memory controller calculates and stores the lengths of
the data groups.
[0108] So long as the host maintains the file of FIG. 13A in an
opened state, a physical write pointer P is also preferably
maintained to define the location for writing any further data
received from the host for that file. Any new data for the file are
written at the end of the file in the physical memory regardless of
the logical position of the new data within the file. The memory
system allows multiple files to remain open at one time, such as 4
or 5 such files, and maintains a write pointer P for each of them.
The write pointers for different files point to locations in
different memory blocks. If the host system wants to open a new
file when the memory system limit of a number of open files already
exists, one of the opened files is first closed and the new file is
then opened. After a file has been closed, there is no longer any
need to maintain the write pointer P for that file.
[0109] FIG. 13B illustrates the appending of data by the host to
the end of the previously written but still open file of FIG. 13A,
also by use of the Write command (FIG. 12A). Data 187 are shown to
be added by the host system to the end of the file, which are also
written in the second block 185 at the end of the data for that
file. The appended data becomes part of the data group (F1,D1),
which therefore now contains more data, since there is neither a
logical nor a physical address discontinuity between the existing
data group 184 and the appended data 189. The full file is thus
still represented as a sequence of index entries (F0,D0), (F1,D1)
in the FIT. The address of the pointer P is also changed to that of
the end of the stored appended data.
[0110] An example of the insertion of a block of data 191 into the
previously written file of FIG. 13A is shown in FIG. 13C. Although
the host is inserting the data 191 into the file, the memory system
appends the inserted data at a location 193 at the end of the file
data previously written. It is not necessary to rewrite the data of
the file in their logical order when data are being inserted into
an open file, although this may be done later in the background
after the host closes the file. Since the inserted data is stored
entirely within the second memory block 185, if forms a single new
group (F1,D3). But the making of this insert results in the
previous data group (F0,D0) of FIG. 13A being divided into two
groups, one (F0,D0) before the insert and one (F2,D1) after the
insert. This is because a new data group needs to be formed
whenever there is a logical discontinuity of the data, such as
occurs at the beginning F1 of the insert and at the end F2 of the
insert. The group (F3,D2) is the result of physical address D2
being the beginning of the second block 185. The groups (F1,D3) and
(F3,D2) are maintained separate, even though they are stored in the
same memory block, because there is a discontinuity in the offsets
of the data stored in them. The original file with the insert is
then represented in the memory system FIT by data group index
entries (F0,D0), (F1,D3), (F2,D1), (F3,D2), in that order. It
should be noted from the examples of FIGS. 13A, 13B and 13C, that
new data for a new or existing file may be written without making
obsolete any data in the memory. That is, execution of the Write
and Insert commands (FIG. 12A) do not cause any other data to be
rendered invalid or obsolete.
[0111] FIG. 13D illustrates another example, wherein a certain
portion of the data originally written in the manner shown in FIG.
13A is updated, using the Update command (FIG. 12A). A portion 195
of the data file is shown to be updated. Rather than rewriting the
entire file in the memory system with the update, an updated
portion 197 of the file is appended to the data previously written.
A portion 199 of the data previously written is now obsolete.
Although it is usually desirable to consolidated the updated file
in order to free up space taken by obsolete data, this is usually
not done while the host maintains the file opened but rather may be
done in the background after the file is closed. After updating,
the file is represented in the memory system FIT by data groups
index entries (F0,D0), (F1,D3), (F2,D1), (F3,D2), in that order.
The single data group (F0,D0) of FIG. 13A is again divided into
pieces in FIG. 13D, one before the updated portion, the updated
portion and one after the updated portion.
[0112] To further illustrate the use of variable length data
groups, a sequence of several write operations involving the same
file is shown by FIGS. 14A-14E in order. The original file data W1
is first written into two blocks of the memory system, as shown in
FIG. 14A, with use of the Write command (FIG. 12A). The file is
then defined by two data groups, the first group starting at the
beginning of a physical memory block and the second group being
required after a physical memory block boundary. The file of FIG.
14A is then described by the following sequence of index entries
for the data groups: (F0,D0), (F1,D1).
[0113] In FIG. 14B, the file data written in FIG. 14A are updated
by use of the Update command (FIG. 12A). Updated file data U1 are
written immediately following the previous group (F1,D1), with the
previous version of the updated data becoming obsolete. Previous
group (F0,D0) of FIG. 14A is shortened to a revised group (F0,D0)
of FIG. 14B, and previous group (F1,D1) is shortened to group
(F4,D2). The updated data are written in two groups (F2,D3) and
(F3,D4) because they overlap a boundary of memory blocks. Some of
the data are stored in a third memory block. The file is now
described by the following sequence of index entries for the data
groups: (F0,D0), (F2,D3), (F3,D4), (F4,D2).
[0114] The file of FIG. 14B is further modified in FIG. 14C by the
insertion of new file data 11, using of the Insert command (FIG.
12A). The new data 11 is written into the memory immediately
following the previous group (F4,D2) of FIG. 14B, as new groups
(F5,D6) and (F6,D7) of FIG. 14C because the inserted data overlap a
boundary of memory blocks. A fourth memory block is used. Previous
group (F0,D0) of FIG. 14B is split into shortened groups (F0,D0)
and (F7,D5) in FIG. 14C, because of the insertion of the new data
11. The file is now described by the following sequence of index
entries for the data groups: (F0,D0), (F5,D6), (F6,D7), (F7,D5),
(F8,D3), (F9,D4), (F10,D2).
[0115] FIG. 14D shows the further modification of the data file of
FIG. 14C that appends new data W2 to the end of the file, using the
Write command (FIG. 12A). New data W2 are written immediately
following the previous group (F10,D2) of Figure 14C, as new group
(F11,D8) of FIG. 14D. The file is now described by the following
sequence of index entries for the data groups: (F0,D0), (F5,D6),
(F6,D7), (F7,D5), (F8,D3), (F9,D4), (F10,D2), (F11,D8).
[0116] A second update to the open file is shown in FIG. 14E, where
updated file data U2 are written to the file of FIG. 14D by the
host issuing an Update command (FIG. 12A). The updated data U2 are
written in FIG. 14E immediately following the previous group
(F11,D8) of FIG. 14D, with the previous version of that data
becoming obsolete. Previous group (F9,D4) of FIG. 14D is shortened
to a revised group (F9,D4) in FIG. 14E, previous group (F10,D2)
becomes fully obsolete, and previous group (F11,D8) is shortened to
form a new group (F14,D9). The updated data are written in new
groups (F12,D10) and (F13,D11) of FIG. 14E, overlapping a block
boundary. A fifth block is now needed to store the file. The file
is now described by the following sequence of index entries for the
data groups: (F0,D0), (F5,D6), (F6,D7), (F7,D5), (F8,D3), (F9,D4),
(F12,D10), (F13,D11), (F14,D9).
[0117] The offsets of the data of each file are preferably
maintained continuous in the correct logical order after the file's
creation or modification according to the preceding description.
Therefore, as part of executing an Insert command, for example,
offsets of the inserted data provided by the host are continuous
from the offset immediately preceding the insert and data already
in the file after the insert are incremented by an amount of the
inserted data. The Update command most commonly results in data
within a given address range of an existing file being replaced by
a like amount of updated data, so the offsets of other data of the
file usually need not be replaced. In place of a separate Update
command, data of a file may alternatively be updated by use of the
Write command since receipt of data from the host with a range of
offsets that already exist in the stored file data can be
interpreted by the memory system as an instruction to update the
data in that offset range.
[0118] It will be noted that all of the data allocation and
indexing functions described above and illustrated by FIGS. 13 and
14 are performed by the controller of the memory system. Along with
one of the commands Write, Insert or Update of FIG. 12A, the host
merely communicates the fileID and offsets of data within the file
that are being sent to the memory system. The memory system does
the rest.
[0119] An advantage of directly writing file data from the host
into the flash memory in the manner just described is that the
granularity or resolution of the data so stored may be maintained
the same as that of the host. If a host application writes file
data with a one-byte granularity, for example, that data may be
also be written into the flash memory with a one-byte granularity.
The amount and location of data within an individual data group is
then measured in a number of bytes. That is, the same offset unit
of data that is separately addressable within the host application
file is also separately addressable within that file when stored in
the flash memory. Any boundaries between data groups of the same
file within a block are then specified in the index table to the
nearest byte or other host offset unit. Similarly, boundaries
between data groups of different files within a block are defined
in the unit of the host offset.
[0120] Although not used to write data, it is relevant to consider
use of the Delete command to delete a portion of a file since such
an operation is converse of the Insert command. The Delete command
can be applied to a range of file offset addresses, with the
following format: Delete <fileID> <offset>
<length>. Data within the portion of the file beginning with
the <offset> and continuing from that address for the
<length> of the deletion are deleted. Offset addresses of the
remaining data of the file after the deletion are then decremented
by the host in order to maintain contiguous offset addresses
throughout the file. This is the converse of the Insert command,
where the host adds data to the middle of a file and then
increments the offsets of the remaining file data after the insert
so that the offsets of the modified file are continuous.
Garbage Collection
[0121] It will be noted from FIGS. 14B and 14E that the Update
command results in the physical space necessary to store the file
being greater than the amount of data in the file. This is because
data that have been replaced by the updates remain stored in the
memory. It is therefore highly desirable to consolidate (garbage
collect) the data of the file into less physical storage space by
eliminating the obsolete, invalid data. More storage space
therefore becomes available for other data.
[0122] It may also be noted that in addition to the file data
updates of FIGS. 14B and 14E, the data insert of FIG. 14C results
in the file data being stored out of order. That is, updates and
inserts are added to the end of the file stored in memory at the
time they are made, while they are nearly always logically
positioned somewhere within the file. This is the case of the
examples of FIGS. 14B, 14C and 14E. It can therefore be desirable
to reorder the data of the file stored in the memory to match the
order of the offsets within the file. This then improves the speed
of reading the stored data since reading the pages and blocks in
sequence will give the data of the file in their offset order. This
also provides the maximum possible defragmentaion of the file. But
reordering the file data to make reading more efficient is not as
important to the performance of the memory system as is file data
consolidation, which potentially frees up one or more memory blocks
for use to store other data. Reordering of the data in a file will
therefore usually not be done by itself, where the benefit is not
worth the added operating overhead, but can be done as part of many
garbage collection operations with little or no added operating
overhead.
[0123] The file of FIG. 14E includes obsolete data groups (gray
portions) stored in the memory because of the two data updates U1
and U2 having been made. The amount of memory capacity being used
to store the file is, as a result, substantially greater than the
size of the file, as is apparent from FIG. 14E. Garbage collection
is therefore appropriate. FIG. 15 provides an illustration of the
result of garbage collecting the data file of FIG. 14E. That file,
before garbage collection, takes up nearly five blocks of storage
capacity (FIG. 14E), while the same file after garbage collection
fits within slightly more than three memory cell blocks (FIG. 15).
As part of the garbage collection operation, data are copied from
the blocks where they are initially written into other erased
blocks, and the original blocks then erased. If an entire file is
data collected, its data may be copied into the new blocks with a
physical order that is the same as the data logical offset order
within the file. The updates U1 and U2, and the insert 11, for
example, are stored after garbage collection (FIG. 15) in the same
order as they appear in the host file.
[0124] Garbage collection also normally results in the formation of
new and different data groups within the file being consolidated.
In the case of FIG. 15, the file is described by the following new
sequence of index entries for the new data groups: (F0,D0),
(F1,D1), (F2,D2), (F3,D3). This is a far fewer number of data
groups than exist with the state of the file shown in FIG. 14E.
There is now one data group for each of the memory cell blocks into
which data of the file have been copied. As part of the garbage
collection operation, the file index table (FIT) is updated to
reflect the new data groups forming the file.
[0125] When a file is a candidate for garbage collection, the FIT
data group entries for that file are examined to determine whether
the file meets set criteria for garbage collection. Garbage
collection will proceed when any of the memory cell blocks
containing data of the file also contain obsolete data. If not,
garbage collection is not necessary. That the file of FIG. 14E
contains obsolete data is apparent from the sequence of data group
index entries given above. For example, it can be seen from the two
successive data groups (F7,D5) and (F8,D3) that obsolete data exist
in the first two blocks where the file data are stored. The
difference in physical address locations D5 and D3 is much greater
that the difference in the logical offsets F7 and F8. It is also
apparent from the file index data group entries that the data have
not been stored in the logical offset order. Alternatively, the
individual FIT entries for a file may retain records of obsolete
data groups of that file that are referenced as obsolete. The
controller then simply scans the FIT entries for each file to
identify any obsolete data groups and the physical blocks in which
they are stored.
[0126] Garbage collection should usually not be performed on an
opened file since the host can continue to make updates or inserts
to the file that generate further obsolete data and/or store the
data out of logical offset order. Many such garbage collection
operations could then result. But in the case where the number of
erased pool blocks falls below a set level, garbage collection of
open files may be necessary to provide enough erased blocks for
storing new data or other operations.
[0127] The closing of a file by the host normally causes garbage
collection to be considered for that file. Garbage collection is
preferably not performed immediately after a file is closed but
rather a closed file is scheduled by the memory controller for
garbage collection in the background when it will not interfere
with current memory operations. A garbage collection queue may be
kept, where a file is added to the queue after closure and then,
when there are no other memory system operations of higher priority
to be carried out, the file that has been in the queue the longest
is selected by the memory controller and garbage collected if that
is needed. Such a selection and garbage collection operation can be
made, for example, in response to receipt from the host of an Idle
command (FIG. 12D).
[0128] Since copying data is the most time consuming part of
garbage collection, particularly for large files, the task may be
divided into components by copying only a portion of the data of a
file at any one time in short bursts. Such partial file copying can
then be interleaved with other memory operations, and even take
place while the host is transferring data to or from the memory
system. The length of the individual copy data bursts can also be
increased in response to the number of erased blocks falling below
some designated number.
[0129] The goal is to perform garbage collection totally in the
background without interfering with or slowing down the primary
operations of transferring data with the host system. But this is
not always possible, particularly if the number of erased blocks
that are available in the erased block pool for programming new
data becomes less than some preset minimum. Garbage collection is
then made a priority and any files in the queue may be garbage
collected in the foreground as a priority in order to consolidate
data so that additional erased blocks are made available for
receiving new data from the host system. If there are no files in
the queue, then open files may have to be garbage collected. When
garbage collection becomes a priority, the host will usually
receive a busy status signal from the memory system and will
therefore suspend any programming of new data until an adequate
number of erased blocks again exist. On the other hand, if there
are a sufficient number or more of erased pool blocks, then the
frequency of garbage collection operations may be reduced, and
garbage collection that could affect performance of the memory
system postponed.
[0130] It will be noted that garbage collection is performed on
host data files. Garbage collection is initiated on a file in
response to the state of the file. When initiated, the index
entries of all the data groups of the file in the FIT are examined
as part of the process. When a file is garbage collected, its data
are copied one data group at a time from their existing block to a
newly opened copy block, in the order specified in the data group
index entries for the file. This is in contrast to existing flash
memory garbage collection, which is based entirely on the status of
individual memory cell blocks.
[0131] However, it a general rule that only the individual blocks
storing data of the file that also contains obsolete data will be
garbage collected. So not all blocks storing data of a file will be
garbage collected. If a file is stored in only two blocks, for
instance, and the first block contains obsolete data and the second
block does not, then the first block will be garbage collected
while data of the second block are left alone. Valid data are
copied from the first block into an erased copy block, and the copy
block will then have some erased pages or other capacity left over
that is in about the amount of the obsolete data. Use of erased
storage capacity less than one block is described below. In the
example of FIG. 14E, the memory space before garbage collection has
obsolete data (gray areas) in each of four of the five blocks
containing data of the file.
[0132] As a modification of the general rule that only those blocks
containing obsolete data are garbage collected, once it is
determined that a given file is to be garbage collected, any data
of the file in a partially filed block is included in the garbage
collection operation. Therefore the U2 data in the fifth bock of
FIG. 14E is included in the garbage collection operation even
though there are no obsolete data in that block. The data in all
five blocks are consolidated together by being copied into four
erased blocks since by not including data of the fifth block, two
of the resulting four copy blocks would be only partially filled
with data. In the case of FIG. 15, only one copy block remains
partially filled. Memory performance is improved by having fewer
partially used blocks.
[0133] Certain file commands of FIG. 12B issued by the host system
may initiate garbage collection. Receipt of the Close command for a
file has already been described. The closed file is placed in the
queue to be garbage collected. The Delete and Erase commands may
also result in garbage collection. The deletion of a file can cause
blocks containing the obsolete file data to be placed in the queue
for garbage collection. The effect of garbage collecting a deleted
file is that there is no valid data of the deleted file, so data
copying to other blocks does not take place. All the blocks
containing only data of the deleted file are simply erased as part
of the garbage collection process. The Erase command has a similar
effect except that the garbage collection of blocks containing only
obsolete data of the erased file may occur immediately or on a
priority basis such as by placing the blocks at the head of the
garbage collection queue.
[0134] A significant advantage of the direct file storage system
being discussed is that the memory learns immediately when the host
deletes a file because the command to do so goes directly to the
memory system. The memory controller can then erase the blocks
storing data of the deleted file as soon as it can be done without
adversely affecting other operations of the memory. The erased
blocks are then made available for the storage of new data.
[0135] In the implementation being described, either of the
commands to Delete or Erase a file causes data of that file to be
erased as a direct response. In a typical logical based interface,
on the other hand, such a command does not reach the memory
controller directly. Rather, the host merely frees up certain
segments of the memory logical address space that were occupied by
the deleted or erased file. Only when the host later writes data to
one or more of those same logical addresses does the memory system
know that the data at the reused logical address segments have been
obsolete. The memory still does not know that data of other logical
address segments occupied by the file have been deleted or erased.
The memory only knows when the host writes data to those other
logical addresses that the data formerly written to those logical
segments have been obsoleted.
Common Blocks
[0136] A result of filling successive blocks with data for a single
file is that when the file is closed and garbage collected, some of
the file data may occupy only a portion of a memory cell block.
Further, data of a small file may not even fill an entire block.
This would result in a considerable amount of the space of
individual blocks going unused if only data of one file could be
stored in a block. Pages of storage capacity within individual
blocks would remain erased and available to store data but would
not be utilized.
[0137] Therefore, some memory cell blocks desirably store smaller
amounts of data from each of two or more files, either entire small
files and/or residual file data left over after other blocks have
been filled. Residual file data are data of a closed file that do
not occupy a full block, and may include one or more data groups. A
file map can keep track of data groups of more than one file in a
single memory cell block in the same manner as if all the data
groups in the block are all of one file. The map of the individual
data groups includes the fileID of which the data group is a part.
A primary purpose of using common blocks is to minimize unused
physical memory storage capacity that can result from writing data
of files into the memory as described above. Most commonly,
residual data of one or more data groups resulting from the garbage
collection of a file will be written into a common block as part of
the garbage collection. An example is the garbage collected file of
FIG. 15, where the last data group (F3,D3) occupies only a small
portion the last block. If it remained that way, most of that last
block would go unused. The data group (F3,D3) of FIG. 15 can be
written into a common block containing one or more data groups from
another one or more files, or that last memory block may be
designated as a common block. In the later case, data groups from
one or more other files would then subsequently be written into
that same block.
[0138] Alternatively, residual file data may be written directly
from the host to a common block during garbage collection when the
amount of residual data is known to be an amount that will fit into
erased space within one of the designated common blocks. The
Close_after command (FIG. 12B) can be used to identify residual
file data and allow it to be written to an open common block that
has enough space to store the amount of data specified as part of
the command, rather than writing the residual file data into an
erase pool block that will not be completely filled. This may
eliminate the need to copy the residual data during a subsequent
garbage collection operation.
[0139] When a new data file is programmed into the memory, the data
may be written into an open common block containing residual data
for one or more other files, beginning at the first available
physical location in the block and proceeding through the locations
of the block sequentially in order, as an alternative to the method
for writing data described previously and illustrated in FIG. 13A.
If the host sends an Open command for a file, followed immediately
by a Close_after command before sending data for the file, it can
be determined that the full data for the file may fit within an
open common block and that the file will be closed before the end
of the open common block is reached. Data for a new file may also
be written into an open common block even if the length of file
data is unknown.
[0140] Therefore, a number of common blocks are desirably
maintained in the memory system for storing common data groups of
two or more different files. A common data group may have a size up
to that of the storage capacity of a block, with a granularity of
one byte. Only one of the common blocks preferably contains data of
a given file but may contain more than one data group of the file.
Also, a common data group is preferably stored in only one common
block, and not divided for storage in multiple common blocks. This
avoids garbage collection in multiple common blocks when the data
group becomes obsolete. A common block is subject to garbage
collection when a data group therein becomes obsolete. Any
remaining valid data group(s) in such a common block are written
into available erased space in another common block or to an erased
pool block, and the common block is then erased. If the common
block being garbage collected contains two or more valid data
groups from different files, they need not be kept together but
rather can be copied to different blocks.
[0141] An example of a common block is shown in FIG. 16, having
been programmed with a common data group from each of three
different data files. Because writing data is constrained in
popular NAND arrays to progress from the pages at one end of the
block to the other, the data groups are stored contiguously with
each other. The white space at the end of the block indicates the
pages of the block that have had no data written into them. It is
difficult to perfectly fit available data groups into the entire
common block.
[0142] The number of blocks designated by the memory system as open
common blocks to which residual data of more than one file may be
written will normally be just a few but may become many if a large
number of files that are garbage collected have residual file data
that does not fit the available space in any existing open common
block. Residual file data are written into one of the open common
blocks that best utilizes the overall memory capacity. When there
is not enough erased space in any of the open common blocks, one of
the existing common blocks is closed and another common block is
opened in its place. Alternatively, an existing open common block
need not be closed, and the number of open common blocks may be
allowed to increase. The new common block may designated as one of
the erased pool blocks but is preferably an incompletely written
block that already contains some residual file data, when
available, such as the last block in FIG. 15 that contains only the
residual file data group (F3,D3). For the garbage collection of a
common block, however, an erased pool block is preferably
designated as a new common block when necessary.
[0143] FIG. 17 illustrates an example process of writing residual
file data in a case where there are five residual or common units
(each a different a block or metablock) that already contain one or
more data groups of residual data from another file. The last data
group (F3,D3) resulting from the garbage collected file of FIG. 15
is an example of such residual file data, although it could include
more than one data group from a single file. There are three
possibilities shown. The first possibility (A) writes the residual
data to residual unit 2 because it has the most erased space
available. The second possibility (B) chooses residual unit 5 for
the residual file data because that is the best fit among the five
residual units. The residual file data nearly fills the unit 5 and
thereby leaves the larger space available in unit 2 for receiving a
larger amount of data in the future. There is not enough space in
any of the units 1, 3 or 4 to take the residual data so these units
are ruled out right away.
[0144] The third possibility (C) writes the residual file data into
a block or metablock unit from the erase pool. Once the residual
file data have been written into the completely erased unit, then
it becomes a residual unit, and can later be opened by the memory
system controller as a common block. In the example shown,
possibility (C) would not normally be used since there is room in
either of the residual units 2 or 5 for the residual file data.
[0145] Alternatively, residual data for a file may be allowed to
split into two or more parts for storage in different common
blocks. For example, no existing open common block may have
sufficient available space for storage of residual data for a
specific file and no erased block may be available to be opened as
a new common block for the residual file data. In this case, the
residual file data may be split between two or more open common
blocks.
[0146] A common block can be closed before it is full of data. As
mentioned above, an open common block with the least amount of
erased space available will generally be closed when it is
necessary to open another common block in order to store a given
amount of residual data of a file. This is in part a consequence of
the preferred operation where residual data for a file are not
split for storage in different common blocks. When the memory
system reports the amount of available memory storage space for new
data to the host by responding to the host's Size command (FIG.
12E), these small amounts of unused capacity in closed common
blocks are not included since they are not immediately useable.
[0147] Because residual file data become obsolete due to the files
of which they are a part being deleted or erased, data within the
common blocks are also consolidated. For example, whenever residual
file data within a common block is designated as obsolete, for
whatever reason, that block is added to either the obsolete file
data block queue discussed above or another queue. If the data
become obsolete because of a host Delete command for its file, then
the common block is placed at the end of the queue. But if it
results from an Erase command, the common block goes to the head of
the queue or otherwise is given priority for garbage
collection.
[0148] Instead of a single queue, there may be the five different
garbage collection queues maintained by the controller during
operation of the memory system, entries in the first two queues
being given priority: [0149] (1) a priority queue of obsolete
blocks, namely those blocks containing only obsolete data as the
result an Erase command (FIG. 12B) for a file; [0150] (2) a
priority queue of common blocks that contain data rendered obsolete
by an Erase command for a file but that also contain some valid
data; [0151] (3) a queue of obsolete blocks (blocks containing only
obsolete data) that result from execution of the Update or Delete
commands (FIGS. 12A and 12B) or because all their valid data have
been copied to another block during garbage collection; [0152] (4)
a queue of common blocks containing some obsolete data but which
also contain valid data, in response to the Delete command or the
discovery during garbage collection that obsolete data exist in a
common block; and [0153] (5) a queue of files for which either the
Close or the Close_after commands have been received. These five
queues may be given priority in the order listed above, all the
items in queue (1) being garbage collected before those of queue
(2), and so on. The blocks are listed in the priority queues (1)
and (2) because all or some, respectively, of their data have been
rendered obsolete by the Erase command rather than by the Delete
command. Blocks and files are added to each of the queues in the
order identified during operation of the memory system, the oldest
in each queue being garbage collected first (first-in-first-out, or
FIFO).
[0154] The priority of items listed in the queues (4) and (5) are
about the same, so could be reversed. Alternatively, there can be a
more complicated priority system where files and common blocks are
selected from the queues (4) and (5) according to a set criteria,
perhaps even before one or more of the higher priority queues are
empty. There are two goals of managing the garbage collection of
common blocks. One is to minimize disruption to normal operation of
the memory system that can result from garbage collection of a
common block when a data group within the common block becomes
obsolete. The other is to minimize indexing updates when a data
group from a common block is relocated during garbage collection of
the block.
[0155] During garbage collection of a common block, all common
groups of valid data remaining in the block are copied one at a
time to one of the common blocks that have the space. If there is
no room in an opened common block for a common group being copied,
the common group is written into a block from the erase pool and
this block may then be designated as a common block. One of the
opened common blocks is then typically closed as part of the
process. As with all garbage collection operations, the file index
table (FIT) is updated to reflect the new storage locations of the
copied data groups.
[0156] The garbage collection queues should be recorded in the
non-volatile memory. A read-modify-write operation on a page or
metapage containing queue information should be performed to add or
to remove an entry. Pages containing garbage collection queue
information may be in a dedicated block or metablock, or may share
a block or metablock with other types of pages, such as a swap
block.
Buffering and Use of Metapages During Programming
[0157] The foregoing describes operation of a memory system without
particular regard to whether the memory cell blocks are linked
together as metablocks or not. The file-based interface between the
host and memory system, and the related memory system operation
described above, work in a memory system using metablocks as well
as in one that does not. The trend is definitely to increase the
amount of data that are written and read at one time (degree of
parallelism) since this directly improves memory performance. The
effective width of the individual memory cell blocks in terms of a
number of data bits is increased by the use of metablocks formed of
two or more such blocks.
[0158] Referring to FIG. 3, for example, all of the memory cells
along each of the word lines within a single block can be
programmed and read together as a page, potentially storing several
one, two, four or more sectors of 512 bytes of user data in each
row. And when two or more blocks are logically linked into a
metablock, all of the memory cells in one row of each of the two or
more blocks can be programmed and read together. The two or more
pages from two or more blocks that are programmed and read together
form a metapage. With the ability to program that much data at one
time, some staging of data within a metapage can be helpful to
fully utilize this available parallelism.
[0159] One metapage of data for each of three host files are
illustrated in FIG. 18, each metapage being shown to include only
two pages for simplicity. The data of each page of a metapage are
programmed into a different block of a metablock.
[0160] For host file 1, the metapage is shown in FIG. 18A to be
filled with data having contiguous offsets (logical addresses), as
a part of a single data group 0. These data are programmed in
parallel into a metapage next in order within a metablock receiving
data for file 1. In FIG. 18B, host file 2 of this example is shown
to be different in that a portion of its first page includes part
of data group 1 with continuous offsets and the remainder of the
metapage contains part of another data group 2 having continuous
offsets. Although there can be a discontinuity in the offsets of
the data where the two data groups join within the file 2 metapage,
these data are programmed together in a single metapage. As
previously described, the direct file storage interface writes data
of a host file as it is received from the host, regardless of the
order of the offsets of the data within the file.
[0161] Some flash memories do not allow the programming of data
into an erased page more than once. In this case, both of the data
groups 1 and 2 of FIG. 18B are programmed at the same time. But if
the memory allows partial page programming, the two data groups of
file 2 can be programmed into a single metapage of the non-volatile
memory at different times. In this case, page 0 would be programmed
at different times, first with data group 1 and then when data
group 2 is programmed into the remainder of page 0 and into all of
page 1. However, it is normally preferred to program both data
groups in parallel, in order to increase system performance.
[0162] In FIG. 18C, the end of a single data group 3 is shown to
have been written into a metapage for file 3. There are certain
conditions when such a small amount of data are programmed into a
metapage of the non-volatile memory while the remaining portions of
that metapage remain erased. One is when the host issues a Close
command (FIG. 12B) for the file of which data group 3 is the
residual file dataFile 3 could also be closed by the memory
controller if the host seeks to open a number of files in excess of
the permitted number and file 3 is chosen for closure in order to
allow another more active file to be opened instead. The memory
controller could also close file 3 if there is insufficient
capacity in the buffer memory for all the file metapage buffers
that have been opened. In any of these cases, it is desirable to
write the partial data contents of a file buffer to the
non-volatile memory right away ("flush" the file buffer). The host
can also send a Shut-down command (FIG. 12D), which means that
power could be lost to the memory and all data within its volatile
buffer memory lost if not programmed into the non-volatile memory
right away.
[0163] If it is later desired to write the beginning of another
data group 4 of file 3 to the memory metapage of FIG. 18C and the
memory does not allow partial page programming, the first page of
data group 4 is written into the second page of the file 3
metapage, as shown in FIG. 18D. This can result in some unused
memory capacity, as indicated by the blank portion between the data
groups 3 and 4 of FIG. 18D, within the non-volatile memory
metapage. This unused capacity may be recovered by a garbage
collection operation performed after file 3 is closed, or may be
allowed to persist since this situation is likely to occur
infrequently.
[0164] It is significant to note that the direct file storage
technique of the present invention can tolerate such unfilled gaps
within blocks of the non-volatile memory. Such gaps cannot easily
be tolerated by current systems, where the memory system interfaces
with the host through a logical address space, as illustrated by
FIGS. 7 and 8. Logical data groups the same size as the memory
blocks or metablocks are mapped into such blocks or metablocks. If
a gap were to be present in the data stored within a block or
metablock of an existing memory system, logical data addresses that
are mapped into the gap would effectively be unusable by the host.
The host assumes in such an existing interface that it has the
entire logical address space available to it but it is very
difficult, if even possible, for new data to be written at logical
addresses that designate the gap.
[0165] A portion of the buffer memory in the system controller,
such as memory 31 of FIG. 2, is typically utilized as programming
data buffers. Buffer capacity in this memory should exist for each
active file being written by the host. Each such "file buffer"
should have capacity equal to at least one metapage in flash
memory. An active file is an open file to which data has been
recently written by the host.
[0166] The number of active host files with which the memory system
deals at any one time may be equal to the number of currently open
memory system files, or may be a lesser number. If the allowed
maximum number of active files is less than the allowed maximum
number of open files, provision must be made for changing the
status of a file between active and open and vice versa. To allow
this, a temporary storage block in flash memory is designated as a
swap block, and data that are less than one metapage in length may
be written from a file buffer to a metapage in the swap buffer, and
vice versa. An index of valid swapped file buffers is maintained in
the swap block. File buffer data are copied to the swap block as an
integral number of pages, followed by a single page providing an
index of the length and location of each copied file buffer, and
the file to which it relates. The swap block is periodically
compacted and written to an erased block when it becomes full.
[0167] When an open file A must be made active as a result of a
host write operation to it, for example, the least recently written
active file B is identified, and its file buffer data copied from
the controller buffer memory to the next available pages in the
swap block. The file buffer data for file A is then identified in
the swap block, and are copied to the available space in the
controller buffer memory previously allocated to file B.
[0168] Data from a file buffer are preferably written to the next
available metapage in the open write block for that file in flash
memory according to one of the cases shown in FIG. 18. In FIG. 18A,
when sufficient data forming part of a single data group 0 within
file 1 is present in the file 1 buffer, it is programmed to a full
metapage in a single operation. In FIG. 18B, when sufficient data
forming the end of data group 1 and the start of data group 2
within file 2 is present in the file 2 buffer, they are programmed
to a full metapage in a single operation. If the flash memory
device supports multiple programming operations on a single page
(partial page programming), when sufficient data forming the end of
data group 1 within file 2 is present in the file 2 buffer, they
may be programmed to part of page 0. When sufficient data forming
the start of data group 2 within file 2 becomes available in the
file 2 buffer, they may be subsequently programmed to the remainder
of the same metapage in a separate programming operation.
[0169] As shown in FIG. 18C, when data forming the end of data
group 3 within file 3 are present in the file 3 buffer, and a
buffer flush operation must be performed, they are programmed to
part of page 0. If the flash memory device does not support
multiple programming operations on a single page (partial page
programming), when sufficient data forming the start of data group
4 within file 3 becomes available in the file 3 buffer, it is
subsequently programmed to page 1 in the same metapage in a
separate programming operation.
[0170] The metapage file buffers are also used during garbage
collection. Valid data groups of a file can be read from the
non-volatile memory and written into the controller buffer metapage
for that file. If the file is being reordered at the same time, the
data groups are written into the buffer in the order of their host
data offsets. The data in each data group, of course, has
contiguous logical offsets. Once the metapage buffer is full, its
data are programmed in parallel into a new block of the
non-volatile memory.
File Indexing
[0171] Each file stored in the memory system is defined by its own
sequence of index entries, as described above, particularly with
respect to FIGS. 13-16. These entries change over time as data for
a file are appended, inserted or updated therein, and when the file
is subjected to garbage collection. FIG. 19 illustrates a sequence
of index entries in a file index table (FIT) for one file at each
of several different times 0, 2, 4 and 5. These are the sequences
described above with respect to FIGS. 14A, 14C, 14E and 15,
respectively. Data in the FIT are preferably written and kept
current by the memory controller without assistance from the host
system. The host system supplies the pathnames, the filenames and
the offsets of data within the file when writing the data to the
memory system but the host does not participate in defining the
data groups or where they are stored in the memory cell array. In
the entries of FIG. 19, the memory cell blocks of FIGS. 14 and 15
are numbered from the left beginning with 1. So for the file in the
state illustrated in FIG. 14C, for example, its third data group
(F6,D7) is noted in FIG. 19 to be stored in block 004, the fourth
block from the left, D7 bytes from that block's initial address.
The length of each data group is also preferably included with each
entry of the table.
[0172] The sequence index entries shown in FIG. 19 for one file are
rewritten by the memory controller when a change to the file
results in a modification to the data groups of the file, or at
other less frequent intervals. The controller can store such
changes in its memory, and then write many of them to the flash
memory at one time. Only one valid set of indices exists for a file
at any one time; four such sets of indices being shown in FIG. 19
that define the file at different times. The memory system
controller then uses the current set of indices of the file as
necessary to program additional data to the file, read data from
the file, garbage collect data of the file, and potentially for
other operations. The FIT is therefore easier to use if the
individual file index entries are stored in order of their file
offsets (Fx) but, if not, the controller can certainly read the
entries in that logical order. The most common cause of the memory
controller reading the index entries for a particular file is in
the course of executing a host command.
[0173] FIG. 20 shows a preferred technique of maintaining and using
the file index table (FIT) as part of a file map. A chain of file
indexing structures is used to identify the physical location of
data with a specified offset address within a file with a specified
pathname. The file map includes a directory 201 that has a logical
structure substantially the same as the standard
disk-operating-system (DOS) root directory and sub-directory
structures used with a conventional logical address interface. The
file directory 201 may be stored in one or more dedicated flash
blocks within the memory system. But the file directory 201 is
preferably managed by the storage system instead of the host. The
host directory commands of FIG. 12C herein are executed by the
memory system controller but at a time and in a manner decided by
the controller. The host is preferably not allowed to write
directly to the file directory.
[0174] Uniformly sized entries, each identifying either a directory
or a file, are kept in the file directory 201. A set of contiguous
entries is used to specify the elements within a specific
directory. A pointer field in an entry specifying a directory
identifies the start of other contiguous entries specifying the
elements within that directory. A pointer field in an entry
specifying a file defines a block number and file number within an
associated file index table (FIT) 203.
[0175] The FIT 203 contains uniformly sized entries of the data
groups, each entry identifying the offset of the start of the data
group and the physical location within the memory system of the
data group. If the logical addresses are maintained with byte
granularity, then the offset of each FIT entry contains a specified
number of bytes from the beginning of the file that designates the
start of the data group. Similarly, the physical location of the
beginning of the data group may be specified with a byte
granularity. The contents of FIT entries for an exemplary file at
different times were described above with respect to FIG. 19. A
separate entry is maintained for each data group. Each file is
defined by a set of contiguous index entries of the data groups
within the file. The number of such entries per file will usually
vary.
[0176] The pathname supplied by the host for a file is used to
traverse the hierarchy of the file directory to obtain the block
number and file number within the FIT. The pathname normally
includes one, two or more directories and sub-directories. These
are used to access the directory within the file directory 201 that
contains a file for which access is sought. The filenames within
that directory are then searched to find an entry 205 for the
filename supplied by the host. When found, the entry 205 provides a
pointer to a group of contiguous entries in the FIT 203 for that
file. These entries are in the nature of those described with
respect to FIG. 19. The offsets of these entries are then compared
with the offset address supplied by the host in order to identify a
correct entry 207, and thus the correct data group. If there is no
identical match of offsets, which may often be the case since the
offsets included in the entries are only the beginning addresses of
the data groups, the entry is selected that identifies a data group
that contains the offset within it. The selected entry 207 of the
FIT 203, in this example, contains the physical block and byte of
the memory location containing data to which access is being sought
by the host.
[0177] The pointer of the directory entry 205 of FIG. 20 for a file
is assigned by the memory system controller when the file is first
created or when the location of the data group entries for the file
in the FIT 203 are changed. This pointer is an example of the
file_handle discussed earlier. In order to avoid the host having to
go through the hierarchy of the file directory 201 in order to find
the FIT entries for a file each time it is accessed, the memory
system may send the file_handle to the host so that it may
thereafter directly access the FIT entries of opened files.
[0178] FIG. 21 illustrates pages within the memory system that
store the individual entries in the file directory 201. Some of the
entries provide pointers to directories within the file directory
201 and others to data within the FIT 203. An example entry 209,
one of many stored in a single memory page, illustrates that each
entry has four fields. The first contains the name of a directory
or file, assigned by the host. The second field contains the
attributes of the directory or file, defined by the host. A pointer
in the third field, in the case of an entry for a directory, points
to another entry in the file directory. In the case of an entry for
a file, the third field contains a pointer to a file entry in the
FIT 203, such as the entry 207. A fourth field, identified as "Data
Groups," is empty in an entry for a directory but, in the case of
an entry for a file, this field specifies the number of data groups
containing data for the file and thus the number of entries in the
FIT 203 for the file.
[0179] It will be noted that the file directory pages illustrated
in FIG. 21 contain two sections. In a most recently written page
211, directory entries (one indicated at 209 in the page 213) and
page pointers both exist. In other directory pages, such as a page
213, directory entries exist but a region of the page contains
obsolete page pointers. Current page pointers are maintained in the
same portion of the most recently written page, in this case the
page 211. One page pointer exists for each logical page in the
directory block, and directs the memory controller to the physical
page corresponding to the logical page it is accessing. By
maintaining the page pointers in the most recently written
directory page, they are updated at the same time as the directory
entry and without a need to then update another page. This
technique is described in more detail for a logical address type of
file system in U.S. patent application Ser. No. 10/917,725, filed
by Gorobets et al. on Aug. 13, 2004.
[0180] In a specific implementation, the directory block contains a
fixed number of logical pages, which is a designated fraction (e.g.
50%) of the total pages in a memory cell block. A directory block
is preferably a metablock dedicated to storage of directory pages
but may be a single erase block. It may have the same parallelism
as metablocks used for storage of data groups, or it may have a
reduced parallelism. When a file directory block becomes full, it
is compacted by copying the valid version of each logical page to a
new block, before erasing the original block. Immediately after a
directory block has been compacted, only a defined fraction (such
as 50%) of the pages in the new copy block are written with
directory entries. The remaining pages are in the erased state to
allow updated or new entries to be written.
[0181] When an entry for a directory or file is created, modified,
or deleted, the set of entries for the directory containing that
directory or file are rewritten in the next available erased page
in the directory block, thus keeping entries for a directory
contiguous. This is done by a read/modify/write operation on the
page previously containing entries for that directory. The previous
page then becomes obsolete. The page pointer entry for that logical
page is also updated to identify its new physical page. This can be
done in the same page programming operation as is used to write the
entries for the directory.
[0182] If the creation of an entry would cause the set of entries
for a directory to overflow a page, it may instead be written as
the first entry in another page. The existing page may also be
rewritten to move the set of entries for the directory to the end
of the page, if necessary. The logical page numbers within the
directory block may be reassigned in the page pointers to keep
entries for the directory contiguous.
[0183] The file index table (FIT) 203 of FIG. 20 stores indices to
the data groups making up all files in the device. The FIT is
stored in one or more dedicated FIT blocks of the memory system.
Each FIT block may store indices for up to a maximum number of
files. The FIT is accessed by means of a logical pointer from the
file directory, specifying one of the files indexed by the table.
The pointer uses indirect addressing by accessing file pointers
provided as part of a most recently written FIT page 217, so that
it does not change as indices are updated and rewritten within a
FIT block.
[0184] The FIT pointer of the file entries in the file directory
201, such as entry 205, has two primary fields. The first is a FIT
block number, identifying one of the logical FIT blocks that makes
up the FIT. The actual physical block address allocated to a FIT
block number is separately managed. The second field of the FIT
pointer is a FIT file number, identifying a logical file number
within the identified FIT block. The logical file number is
translated to a physical file entry location by a specific file
pointer stored within the most recently written page of the FIT
block.
[0185] The FIT does not have a predefined size, and the number of
FIT blocks is a function of the number of files and the number of
data groups into which the data is organized. New FIT blocks will
be created and FIT blocks eliminated during operation of the
device.
[0186] With reference to FIG. 22, the individual FIT pages have a
number of data group entries, one entry 219 for each data group, as
previously described with respect to FIG. 19. In this specific
example, each entry 219 contains four fields. The file offset field
specifies the offset address within the file of the start of the
data group being identified by the entry. The block address field
specifies the physical address of the block or metablock within the
memory system that contains the data group. The byte address field
specifies the address of the page within the block or metablock,
and the byte within the page, at which the data group starts. The
fourth field is the length of the data group. The data group length
may not be necessary in all cases since the length of a data group
can be calculated from data of adjacent entries. By maintaining the
length, however, it is not necessary to make this calculation each
time the length of a data group is desired. This is particularly
valuable information if the data group entries are not written in
the FIT in the order of their logical addresses, where the
calculation then becomes more difficult.
[0187] A host file is preferably defined by a set of contiguous
data group entries that together define the ordered data groups
that form the file. The shaded contiguous group entries shown in
the page 215 of FIG. 22 define one file, and those in the page 217
another file. For each of the individual files in the FIT, there is
one file pointer 221 in the last written FIT page 217 for the
individual files maintained in that block. The file pointer 221
defines the start of the file entry for a file with a specific FIT
file number within the FIT block. It contains two fields. One field
is the physical number of the page where the data group index
entries reside, and the other field is the number of the first
group entry within that page. A file pointer exists for each
possible file number in the FIT block, i.e. the number of file
pointers is equal to the maximum number of files in a FIT block. A
reserved code (not shown) in a file pointer entry indicates that
the specific file number is unused in the FIT block. Although file
pointers may exist in other FIT pages, they are valid only in the
most recently written page of the FIT block.
[0188] A FIT block is preferably a metablock dedicated to storage
of FIT pages. It may have the same parallelism as metablocks used
for storage of data groups, or it may have a reduced parallelism. A
single erase block may be used instead. Immediately after a FIT
block has been compacted, only a defined fraction (such as 50%) of
the pages in the block should contain entries. The remaining pages
should be in the erased state to allow updated or new entries to be
written. More than one FIT block may be necessary when a large
number of files are being maintained.
[0189] The FIT is updated by writing the data group entries of one
or more complete files in the next available unprogrammed page of
the FIT block. The file pointer entries for the files are also
updated to identify the new locations of the file entries. Both
file data group entries and the file pointers are written in the
same page program operation. The previous location of the file
entry then becomes obsolete in the FIT block. Alternatively, a file
entry may be updated by writing it in a different FIT block or a
new FIT block. In this case, the file pointer in both blocks should
be updated, and the logical pointer for the file in the file
directory should be modified.
[0190] When a FIT block is full, valid group entries are copied in
a compacted form to a new erased block and the previous FIT block
is erased. The logical FIT block number and the logical file
numbers are not changed by this operation. The number of data group
entries should be restricted such that only a defined fraction
(such as fifty percent) of the pages are programmed in the
compacted FIT block. File entries should be moved to other FIT
blocks, if necessary, and their logical pointers in the file
directory modified.
[0191] It is desirable to keep all the FIT entries of an individual
file in one page or metapage. This makes it relatively easy to read
all the entries for a file. Although the FIT entries of a file
could be chained together by including in each the physical address
of the next, this would likely require reading more than one FIT
page or metapage. It is also desirable to keep the FIT entries
contiguous within the page. This can result in frequently writing
new pages as the number of FIT entries increases, particularly when
a closed file is opened and data groups are added to it. As new FIT
entries are written into a new page, the existing entries are
copied from another page and written along with the new entries in
the new page. The space taken by the obsolete data in the prior
page of the existing entries is retrieved for use when the FIT
block is compacted.
[0192] The individual data groups, as discussed above, are
contained within a single block or metablock, may start on any byte
boundary within the block, and may have a length of any integral
number of bytes. A header may be added to each data group by the
controller as it is written into the memory. Such a header may
include two fields. A first field contains a code to identify the
start of a data group. This code may be the same for all data
groups and is chosen as a bit pattern that infrequently exists in
the data stored as part of the groups but it is not necessarily
that the code never appear in such data. The start of a data group
can be found by the memory controller scanning the stored data for
this code and then confirming when it finds the bit pattern of the
code that it is indeed the start of a data group and not data
within a group. The second field of the header allows this
confirmation to be made. This second field is a pointer to the file
entry in the FIT 203 (FIGS. 20 and 22) for the file containing the
data group. The second field defines a FIT block number and FIT
file entry. The controller then reads the FIT file entry to see if
it points back to the data group from which the code was read. If
so, the code is confirmed to be an indication that it is within a
header of a data group. If not, the bit pattern of the code is
known to have been read from other data so is ignored as a data
group header.
[0193] By including such a header as part of the data stored in the
individual data groups, the file to which any data group in the
memory belongs may be determined by reading its header and
accessing the FIT entry to which the header points. The FIT entry
219 (FIG. 22) could include the fileID. This capability is used
herein, for example, to identify data groups in a common block
during garbage collection of the common block. If garbage
collection is instituted because one or more other data groups
within the common block are obsolete, it is then desirable to be
able to identify the file to which each remaining valid data group
belongs.
[0194] Instead of a header being made part of each data group, one
header can be provided per file. It is advantageous to be able to
identify the full pathname for a file when the location of a FIT
entry for a data group in the file is known, and a file header
makes this possible. Such a header may be provided as the first of
the set of entries in the FIT for a file, or the first of a set of
entries for a directory in the file directory.
[0195] This file header contains a code to identify it as a FIT
file header, and a reverse pointer to the directory entry that is
pointing to it. The FIT file header may also contain additional
information, such as the length of the file and the number of data
groups. Alternatively, the header entry may occupy the space of one
or more normal entries.
[0196] Similarly, a directory header entry contains a code to
identify it as a directory header, and a reverse pointer to the
directory entry that is pointing to it. The root directory is
identified explicitly in its header. The directory header may also
contain additional information, such as the number of entries in
the directory.
[0197] As an alternative to use of such data group or file headers,
an index may be written at the end of each common block to identify
the file to which each data group in the common block relates. When
a common block is closed, such an index may be written. The index
preferably contains a physical byte address within the common block
where each data group in that block starts, and a pointer to the
FIT entry for that data group. The file to which each common block
data group is a part may then be determined by reference to the FIT
entry provided in the data group's index entry.
[0198] The data group header described above may still be retained,
even if these common block indices are employed, if it is desired
to have some redundancy of file identification for the data groups
stored in the memory.
[0199] Since the memory device manages the file directory and file
indexing, it may therefore control device configuration parameters
that are reported to a host. This is not normally possible when the
file system for a device is managed by the host, as in current
commercial systems. The memory device may vary the reported storage
capacity, for example, both the capacity of the total device and
the available unwritten capacity.
[0200] The pathname that is used for files at the Direct-File
Interface may identify the ID of the storage device itself, and a
partition within the storage device. It may be an advantage to
allow the memory device to modify the size and number of
partitions. If a partition becomes full, unused capacity from one
or more other partitions may be reassigned to the full partition.
Similarly, unused capacity from other partitions may be assigned if
a new partition is created. The device may modify reported capacity
for the various partitions, as above.
[0201] The directory structure shown in FIGS. 20 and 21 uses a
pointer in a directory entry to identify the start of a set of
other contiguous entries for elements within that directory. This
pointer points to a logical page within a directory block, and
remains unchanged when the physical page corresponding to that
logical page is changed. However, the number of elements within a
directory is frequently changed, and the set of entries for a
specific directory will frequently need to be moved from one
logical page to another, or be moved within its current logical
page. This can result in a frequent need to update pointer
references within the directory blocks. An alternate entry indexing
technique for the directory structure to that of FIG. 21 is shown
in FIG. 23, where its corresponding elements are identified with
the same reference numbers but with a prime (') added.
[0202] The structure of FIG. 23 is the same as that of the FIT of
FIG. 22 but with different terminology that relates to directories.
Specific entries are pointed to rather than just a page as done in
FIG. 21. This provides a more efficient method of indirect
addressing within a directory block for a set of entries for a
specified directory than that described above with respect to FIG.
21. It is appropriate to use the same indexing scheme for directory
(FIG. 23) and FIT (FIG. 22) blocks, because both have very similar
characteristics and requirements. They each store sets of
contiguous entries, relating to either directories or files. The
updating of a set of entries may consist of either changing the
content of existing entries, or changing the number of entries.
[0203] The method described with respect to FIG. 22 for updating
FIT entries leaves a fraction (e.g. 50%) of the pages in a physical
block containing the FIT entries in the erased state after
compaction of the block. This then allows for updated FIT entries
to be written into the remainder of the compacted block. All FIT
blocks incorporate this capacity overhead, even when no files
indexed by a block are actually open. The FIT therefore occupies
more memory capacity than is desirable.
[0204] An alternative FIT update method causes updated sets of FIT
entries to be written in a separate FIT update block, rather than
into available erased capacity in the blocks in which they were
originally located. FIGS. 24 and 25 outline indexing techniques for
the file directory and the FIT, respectively, that utilize update
blocks. The techniques are identical, with only terminology
differences for the directory and the FIT.
[0205] The description below relates to updating group entries in a
FIT block as shown in FIG. 25 but is equally applicable to updating
group entries in a file directory (FIG. 24).
[0206] Reading group entries from the FIT block is as described
above. The FIT Block No field of the FIT Pointer defines the
logical FIT block. A FIT Block List is contained in a data
structure in flash, and provides the information to translate the
FIT Block No to the physical address of the block in which it is
located. A block address in the FIT Block List is updated whenever
a FIT Block is moved during a compaction or consolidation
operation.
[0207] The File Pointers region of the most recently written page
in that block allows the FIT File No value in the FIT Pointer to be
translated to a File Pointer designating the start of the set of
File Group Entries for the specified file. The File Group Entries
may then be read from the FIT Block.
[0208] When the content of a group entry, or the number of group
entries in the set for a file, is updated, the complete set of
entries is rewritten in an erased page. (This is based on as
assumption that a page being the minimum unit of programming in the
flash memory, with multiple write operations to the same page
prohibited between erase operations.) The set may occupy multiple
pages, if required.
[0209] In the technique currently being described, this page is the
next available page in the FIT block. In the revised scheme, this
page is the next available page in a separate FIT Update Block. A
FIT update block has the same page structure as a FIT block, as
shown in FIG. 23. Its existence is identified by the presence of
the target FIT Block No in a FIT Update Block List, which also
contains the physical block address for the update block and an
Update File No corresponding to the original FIT File No. There may
be a separate FIT Update Block for each FIT Block being updated, or
preferably a FIT Update Block may relate to multiple FIT Blocks. A
single FIT Block may also relate to multiple FIT Update Blocks.
[0210] When a FIT Update Block becomes filled, its valid data may
be written in compacted form to an erased block, which becomes a
new FIT Update Block. There may be as little as a single page of
valid data, if updates have related to only a few files. Multiple
FIT Update Blocks may be compacted together to a single block.
Block compaction is preferred to block consolidation if the update
block related to a file or files that are still open, which may
continue to be updated.
[0211] When group entries for a file are updated in a FIT Update
Block, the entries for that file in the original FIT Block become
obsolete. At some stage, the original FIT Block must undergo a
garbage collection operation to clean it up. This may be done by
consolidating valid data in the FIT Block and FIT Update Block into
an erased block.
[0212] If the number of entries has increased during the update
process, and valid data cannot be consolidated into a single erased
block, files originally assigned to that FIT block may be
reassigned to two or more FIT blocks, and the consolidation may be
performed to two or more blocks.
[0213] Entries from a FIT Update Block may be consolidated with
entries from a FIT Block, and therefore eliminated from the FIT
Update Block, whilst entries for other files may remain in the FIT
Update Block.
[0214] As a further alternative, the Directory Block and FIT Block
structures may be merged into a single Index Block structure, which
may contain both directory entries and file group entries. An Index
Update Block may act as an update block for both directory entries
and file group entries, either when there are separate Directory
Block and FIT Block structures, or when there is a combined Index
Block structure.
[0215] It will be recognized that the file directory and FIT
described above with respect to FIGS. 20-25 are modeled on DOS
systems. Alternatively, they may be modeled on Linux, Unix, the NT
File System (NTFS) or some other known operating system.
Specific Memory System Operation Example
[0216] Operational flowcharts of FIGS. 26-32 provide an example of
the operation of a memory system constructed as described above
with respect to FIGS. 2-6 but with use of a specific combination of
the direct file interface techniques described with respect to
FIGS. 9 and 10-22. The functions included in the flowcharts of
FIGS. 26-32 are carried out primarily by the controller 11 (FIG. 2)
executing its stored firmware.
[0217] Referring initially to FIG. 26, an overall system operation
is shown. In a first step 251, the memory system is initialized,
which includes the processor 27 (FIG. 2) executing boot code in the
ROM 29 to load firmware from the non-volatile memory into the RAM
31. After initialization, under control of this firmware, the
memory system then looks for a command from the host system, as
indicated by step 253. If a host command is pending, the command is
identified from among those listed in FIG. 12. If a Read command
(FIG. 12A), it is identified in a step 255 and executed by a
process indicated at 257 that is more fully described below with
respect to FIG. 27. If not a read command, any one of the Write,
Insert or Update programming commands of FIG. 12A is identified at
a step 259 of FIG. 26 and executed by a process 261 that is the
subject of FIG. 28. If a Delete or Erase command (FIG. 12B), it is
identified in a step 262 and executed in a step 263, as described
in more detail in FIG. 30. An Idle command (FIG. 12D) from the host
is identified at 264 of FIG. 26 and results in a garbage collection
operation 265 that is shown in a flowchart of FIG. 31. This example
of FIGS. 26-32 is described for a memory system operating with
metapages and metablocks, as previously described, but can also be
executed by a memory system organized in pages and/or blocks.
[0218] If the host command is other than Read, Write, Insert,
Update, Delete, Erase or Idle, in this specific example, such
another command is executed by a step 267 of FIG. 26. Among these
other commands are those that cause a garbage collection operation
to be added to a queue, such as the Close and Close_after commands
listed in FIG. 12B and described above. After executing a received
command by any of the steps 257, 261, 263, 265 or 267, a next step
268 inquires whether the priority garbage collection queues are
empty or not. If so, the processing returns to the step 253 to
execute a pending host command, if one exists. If not, the
processing returns to the step 265 to continue garbage collection
rather than allow the possibility of another host command to be
executed. For the specific example described with respect to the
flowcharts of FIGS. 26-32, there are the five different garbage
collection queues described above: two priority queues for obsolete
metablocks (metablocks with only obsolete data) and common
metablocks with obsolete data, wherein the obsolete data result
from an Erase command; two other queues for obsolete metablocks and
common metablocks with obsolete data, wherein the obsolete data
result from other than execution of Erase commands; and one queue
for files to be garbage collected. For garbage collection listed in
the three non-priority queues, another pending host command is
given priority to carrying out the listed garbage collection
operations. However, for those in the priority queues, garbage
collection is given priority over execution of a new host command.
That is, the host can be made to wait until any priority garbage
collection operations are completed before a new host command may
be executed. This is because the host has previously given priority
to erasure of data in the metablocks in the priority queues by use
of the Erase command. The priority garbage collection also results
in additional erased metablocks being generated with relatively
little processing time. But if not priority, garbage collection is
performed in the background, when the host is idle, or interleaved
with other operations when necessary to maintain the erased block
pool. If it is determined by the step 268 that there are no
priority garbage collection operations pending, the next step is
then step 253, where a new host command, if any, is executed.
[0219] Returning to the step 253 of FIG. 26, if no host command is
pending, garbage collection 265 is carried out, including
non-priority garbage collection, if it is determined by a step 269
that the host has been inactive or idle for a predetermined length
of time, or that the most recently received host command was the
Idle command, or by the step 264 that the pending host command is
the Idle command. Under these circumstances, the garbage collection
is then most likely performed entirely in the background.
[0220] Execution of a Read command in the step 257 of FIG. 26 is
the subject of the flowchart of FIG. 27. A first step 271 is to
read the file index table (FIT), as described above with respect to
FIGS. 19 and 22. The host Read command includes the fileID and the
offset within the file where reading is to start (see FIG. 12A).
All of the FIT entries for the file to be read, or portions
thereof, are preferably read from the non-volatile memory into the
controller memory 31 (FIG. 2) in order to avoid the necessity to
read the FIT from the non-volatile memory each time some data are
needed from it. In a step 273, a data group number counter is
initialized to the number of the data group making up the requested
file where the starting offset lies. This is done by comparing the
host specified starting offset with those of the data groups in the
FIT entries for the host designated file. Next, in a step 275, a
data group length counter is initialized to the amount of data
within the initial data group from the host-supplied offset to the
end of the data group. One data group is read at a time, and steps
273 and 275 set up two counters that manage the reading of the
first data group. The beginning physical byte address within the
non-volatile memory of the data to be read is determined from the
FIT entry for the initial data group. The logical offset and
physical byte address of data within a group are linearly related,
so the starting byte address, if reading is not starting at the
beginning of the data group, is calculated from the host supplied
beginning offset within the file. Alternatively, in order to
simplify initialization of the data group length counter, the
length of the each data group can be added to its record 219 of the
FIT (FIG. 22).
[0221] It is assumed for the rest of this description of the
flowcharts that the memory system is operating with metablocks and
data are read and programmed in units of metapages, as previously
described. A data metapage is read in a step 277 from the initial
data group stored in the non-volatile memory that includes the
starting byte address. The read data are typically then written
into a controller buffer memory (RAM 31 of FIG. 2, for example) for
transfer to the host. The data group length counter is then, at a
step 279, decremented by the one metapage. That counter is then
read in a step 281 to determine whether it has reached zero. If
not, there are more data of the initial data group to be read. But
before returning to the step 277 to read the next metapage of data
in order, a step 283 checks whether the host has issued another
command. If so, the reading operation is terminated and the process
returns to step 253 of FIG. 26 to identify the received command and
then execute it. If not, reading of the initial data group
continues by reading its next metapage in steps 277 and 279 of FIG.
27. This continues until the data group length counter is
determined by the step 281 to have reached zero.
[0222] When this occurs, the FIT entries for the file are again
read, in a step 285, to determine, in a step 287, whether there are
any further data groups of the current file to be read. If so, the
data group number counter is updated in a step 289 to identify the
next data group, and the data group length counter is initialized
in a step 291 to the length of the data in the new group. The first
metapage of the new data group is then read in a step 277, if it is
determined by the step 283 that no other host command is pending.
The steps 277 and 279 are then repeated for each metapage of the
new data group until all of its data are read, at which time the
steps 285 and 287 determine whether yet another data group exists,
and so on. When it is determined by the step 287 that all the data
groups of the file after the host supplied offset have been read,
the processing returns to the step 253 of FIG. 26 to execute
another host command.
[0223] In a special case where a file is being used as a circular
buffer, reading of the file may be repeated after the step 287 of
FIG. 27, rather than returning to the step 253 of FIG. 26. This can
occur in the case where data of the current file are being
programmed by the host during reading by responding in the step 283
to a host command to write data to the same file.
[0224] An example of the data programming operation 261 of FIG. 26
is given in the flowchart of FIG. 28. When one of the data
programming commands of FIG. 12A is received from the host, that
command includes a fileID of the file to which data are to be
written. A first step 295 determines whether that designated file
is currently opened for programming. If so, a next step 297
determines whether there are data in the controller buffer (such as
RAM 31 of FIG. 2) for that file. Data are transferred by the host
system into the controller buffer memory, and then by the memory
system controller into the flash memory.
[0225] But if the host designated file is determined by the step
295 not to be opened, a next step 299 asks whether the number of
files currently opened by the memory system for programming is
equal to or greater than a maximum number N1 that is allowed by the
memory system to be open at the same time. The number N1 is preset
within the memory system, and can be 5, 8 or some other number of
files. If the number of opened files is less than N1, a next step
301 causes a new file to be opened by providing system resources
necessary to program data to that file, and the processing proceeds
to the step 297. If the number of open files is determined in the
step 299 to be equal to or greater than N1, however, a currently
opened file first needs to be closed, as indicated by a step 303,
before the new file can be opened in the step 301. The basis upon
which an open file is selected to be closed in the step 303 can
vary but most commonly will be the opened file to which data has
least recently been written by the host. It is presumed from this
that the host is unlikely to write data to this file in the near
future. But if it does, then that file is reopened after another
open file is closed, if necessary at that time.
[0226] When it is determined in step 297 that at least a metapage
of data of the current file are in the controller buffer, a next
step 305 determines whether a metablock within the memory array has
been opened for programming. If so, data are then programmed from
the controller buffer memory into the open metablock, in a step
307. If not, a metablock is first opened, in a step 308, by
providing the system resources necessary to program data to that
metablock. Data are written into the open memory metablock, in this
example, in units of one metapage at a time. Once this unit of data
is written, a next step 309 determines whether the opened write
metablock is full of data. If not, the process normally proceeds
through steps 311 and 313 and back to the step 297 to repeat the
process for programming the next metapage of data in the currently
opened file.
[0227] If the write metablock is determined by the step 309 to be
full, however, that metablock is closed, in a step 315, and the FIT
is updated, in a step 317, including closing the current data group
since a memory metablock boundary has been reached. Metablock
entries of the lower priority obsolete metablock garbage collection
queue are then updated, in a step 318. During the FIT update in the
step 317, it is determined whether filling the write metablock has
created another metablock that contains all obsolete data of the
current file or is a common metablock containing obsolete data of
the current file. If so, that metablock is added to an appropriate
lower priority metablock queue by the step 318 for garbage
collection. The processing then returns to the step 311 and passes
through the step 313 back to the step 297. This time through the
steps 305 and 307, the step 308 will open a new write metablock
since the previous metablock has just been closed by the step
315.
[0228] The step 311 asks, after each metapage of data have been
written by a path including the step 307, whether the number of
metablocks currently existing in the erased metablock pool is in
excess of a minimum number N2 that has been determined necessary to
efficiently operate the memory system. If so, the step 313 asks
whether another host command has been received. If no other host
commend is pending, the step 297 is repeated for the next metapage
of data to be programmed into the memory. But if a host command has
been received, the FIT is updated, in a step 319, to close the data
group that has been written. After updating the metablock entries
in a lower priority obsolete metablock garbage collection queue in
a step 320 (similar to the step 318 described above), the process
returns to the step 253 of FIG. 26.
[0229] But if it is determined in the step 311 of FIG. 28 that
there is a shortage of erased metablocks for the storage of data
(equal to or less than the preset number N2), a sub-routine is
executed to garbage collect a file or common metablock in the
foreground, in order to increase the number of erased metablocks.
Such garbage collection is preferably not performed after each
metapage of data are written into the memory but rather only after
each N3 metapages have been written. A write metablock metapage
program counter maintains a count of a number of metapages of host
data that have been programmed in succession without any garbage
collection in between them. This counter is reset to zero each time
garbage collection is performed and then incremented by one each
time a metapage of data are programmed. In step 321, it is
determined whether that count is in excess of the predetermined
number N3. If not, the counter is incremented by one, in a step
323, to record that a metapage was written into the memory in the
step 307. But if the count of the program counter is in excess of
the number N3, then garbage collection takes place by steps 325,
327 and 329, followed by the program counter being reset to zero in
step 331.
[0230] The purpose of the garbage collection at this point in the
process is to create additional erased metablocks for the erased
metablock pool. But only a portion of a garbage collection
operation is preferably performed each time by execution of the
steps 325, 327 and 329. The copy operation is divided into smaller
operations that are spread out over time in order to minimize the
duration of the intervals that the memory is unavailable to the
host, thereby to minimize the adverse impact on data programming
performance. And the partial garbage collection operation is
performed only every N3 programming cycles for that same reason.
Although the numbers N2 and N3 can be preset in the system, they
may alternatively be determined during operation of the memory
system to adapt to particular conditions that may be
encountered.
[0231] Since most of the time required by a complete garbage
collection operation on a file or common metablock to provide one
or more additional erased metablocks is taken by copying valid data
into one or more copy metablocks, it is this copying that is
primarily interleaved in FIG. 28 with the data programming. A step
325 selects a foreground mode for the garbage collection 329, which
is described with respect to the flowchart of FIG. 31. A certain
preset number of metapages of a data group are then copied in
succession from a metablock being garbage collected and into a
previously erased copy metablock. The step 327 resets a counter of
the metapages copied in this interleaved fashion, so that in the
step 329 the preset numbers of metapages are copied before the
operation returns out of the garbage collection step and back into
the data programming loop. The write metablock metapage program
counter referenced by the step 321 is then reset by a step 331
after each garbage collection operation 329. This causes N3
metapages of data to be written into the memory before another
garbage collection operation takes place in the foreground, and
thereby spreads out any delays in data programming that are caused
by the garbage collection.
[0232] The interleaved nature of this garbage-programming algorithm
is illustrated by a timeline of FIG. 29. N3 metapages of data from
the host are written consecutively into the memory between garbage
collection operations. Each garbage collection operation is limited
to copying N4 metapages of data, after which programming of another
N3 metapages of data takes place. While in the garbage collection
phase, the host may be caused to wait for the memory system to
complete it before transferring additional metapages of data to the
memory system controller buffer.
[0233] Execution of the delete file routine 263 of FIG. 26 is shown
by the flowchart of FIG. 30. Its primary purpose is to identify
obsolete blocks and common blocks with obsolete data of the file
being deleted, and then place those blocks in the appropriate
garbage collection queues. When either a Delete or Erase command
(FIG. 12B) is received by the memory system along with the name or
other identification of the file to be deleted, a step 335
initializes a data group number counter to zero. In a step 337, the
file directory and file index table (FIT) are read to obtain the
identities of the data groups stored in the memory that make up the
designated file. Each data group of the file is then examined one
at a time by incrementing the data group number counter that points
to each data group in logical sequence.
[0234] A first inquiry 339 for a data group is whether it lies in a
common metablock or not. If so, a step 341 adds that metablock to
one of the common block garbage collection queues. The metablock is
placed in the priority common metablock queue if the file is being
deleted by an Erase command, and in the other common metablock
queue if by the Delete command. Any common metablock with a data
group of a file to be deleted is scheduled for garbage collection.
If it is determined by the step 339 that the data group is not in a
common metablock, a next inquiry 343 determines whether the data
group is within a metablock already scheduled for garbage
collection by being in the obsolete metablock garbage collection
queue. If the metablock is already scheduled for garbage
collection, it should not again be added to that same queue. But if
not so scheduled, such it is added by a step 345 to one of the
obsolete metablock garbage collection queues. The metablock is
placed in the priority obsolete metablock queue if the file is
being deleted by an Erase command, and in the other obsolete
metablock queue if by the Delete command.
[0235] After either of the steps 341 or 345 has been executed, or
the inquiry of the step 343 is positive, the process for one data
group is completed. A next step 347 increments the data group
number counter. An inquiry 349 is then made as to whether there is
another data group in the file. If so, the processing returns to
the step 339 and is repeated for that next data group. If not, it
is then known that all metablocks containing data of the file to be
deleted have been entered into the obsolete and common metablock
garbage collection queues. The FIT is then updated, in a step 351,
to obsolete the data group records of the file to be deleted.
Records in the FIT of obsolete data groups are then typically
eliminated when the FIT is compacted. In a final step 353, the file
directory 201 (FIG. 20) is updated to remove the deleted file from
it.
[0236] The flowchart of FIG. 31 illustrates a specific example of
the garbage collection operations 265 of FIGS. 26 and 329 of FIG.
28. This algorithm is entered in the background when the host sends
an Idle command (step 264 of FIG. 26) or the host has been idle for
a time (step 265 of FIG. 26), or in the foreground during a
programming operation (step 329 of FIG. 28) when fewer than N2
erased metablocks remain in the erased metablock pool. A first
inquiry 355 is whether there is a garbage collection operation in
progress that has not yet been completed. It will be seen from this
description of FIG. 31 that the processing comes out of garbage
collection under certain circumstances, such as when a host issues
another command or when data copying is interleaved with other
operations. So if there is an uncompleted garbage collection
operation pending, a step 357 is next since a pending garbage
collection is given priority. But if it is determined by the step
355 that there is no pending garbage collection operation, a next
step 356 then looks at the various garbage collection queues to see
if there is at least one entry in them. If so a next step 358
selects one of the entries, if there are more than one, in
accordance with the priorities discussed above. Garbage collection
of a next in order metablock in the obsolete metablock block queue
will generally be given priority over a metablock on the common
metablock queue or a file on the queue of files to be garbage
collected. This is because the size of the erased metablock pool
can be increased more quickly by means of garbage collecting
obsolete metablocks since no data need be copied.
[0237] A next set of steps 360, 362 and 366 determine the object of
the garbage collection for the selected queue entry, since the
process is different for files, common metablocks and obsolete
metablocks. The inquiry 360 asks whether it is a file that is being
garbage collected. If so, the FIT entries for the file are read in
order, in steps 370 and 372, to set certain counters and a count.
In the step 370, a first counter is set to the number of data
groups in the file and a second counter is set to the number of
metapages (length) of data in the first data group of the file.
This length may be calculated from the FIT data group entries if
their lengths are not part of their entries. Alternatively, the
length of each data group may be included as part of its entry in
the FIT, as shown in FIGS. 20 and 22, in order to eliminate the
need to calculate a data group length each time it is needed. In
the step 372, a third counter is set to the number of metapages of
current file data stored in memory blocks containing obsolete data
or data from other files. These are the data of the current file
that need to be moved. This third count preferably ignores all
obsolete and invalid data groups within the file. Finally, a count
is made of a number of residual metapages of data that would result
if the valid data of the file were compacted into and filled an
integral number of metablocks. That is, the residual metapage count
is an amount of data of the file, less than a metablock, that would
have to occupy a common or partially filled metablock should all
the metablocks containing data of the file be included in the
garbage collection. After these three counters are set and the
residual metapage count made and stored, a next step 359 begins
garbage collecting data of the file.
[0238] Returning to the step 360, if the entry selected by the step
358 is not a file, a next inquiry 362 determines whether it is a
common metablock that is to be garbage collected. If so, common
metablock garbage collection 364 of FIG. 32 is carried out. If not,
an inquiry 366 determines whether the selected entry is for an
obsolete metablock, such as one added to an obsolete metablock
queue by the step 345 of FIG. 30. If so, the selected metablock is
erased in a step 368. After the common metablock garbage collection
364, the metablock erasure 368 or if the inquiry 366 results in a
negative response, the processing returns to the step 253 of FIG.
26. If the garbage collection operation remains incomplete, then it
will be resumed the next time the garbage collection algorithm is
entered.
[0239] Returning to the step 355, if there is a pending garbage
collection, that will be further processed. The step 357 of FIG. 31
determines whether the pending operation is for a file or not, like
the step 360 for a new item selected from a garbage collection
queue. If not for a file, the process of the steps 362, 364, 366
and 368, described above for a new item, are executed and garbage
collection then ended by returning to the step 253 of FIG. 26.
[0240] If it is determined by the step 357 of FIG. 31 that the
resumed garbage collection is for a file, the inquiry 359 is made
next. The processing that follows is substantially the same for a
file for which garbage collection is resumed (through the step 357)
or for a new file selected from a queue (through the step 360) for
which its counters and residual metapage count are set by the steps
370 and 372. When garbage collection of a file is in progress and
being resumed through the step 357, the data group number and file
metapage number counters and the residual metapage count were set
when the garbage collection of the file was first begun. The data
group number and metapage counters have likely been decremented to
different values than originally calculated by the earlier partial
garbage collection of the current file. All four are stored when
garbage collection of a file is suspended and then accessed during
the resumed processing of the same file.
[0241] The common step 359 inquires whether there are one or more
metapages within the current data group that have yet to be copied.
This is determined by reference to the data group length counter.
One metapage of a valid data group is examined and copied at a time
by the subsequent steps. If data are determined by the step 359 to
remain in a pending data group, a next inquiry 361 determines
whether a copy metablock is open. If so, a metapage of data of the
current data group of the file being garbage collected is read from
the metablock in which it is stored, by a step 363, and written
into the copy metablock, in step 365. The data group length counter
is then decremented in a step 367 by one metapage, and the file
metapage counter is decremented by one in a step 369.
[0242] If it is determined by the step 361 that a copy metablock is
not opened, a next step 371 determines whether the file being
garbage collected is an open file. If so, a next step 373 opens a
copy metablock, and the process then proceeds to the step 363
previously described. But if not an open file, a next step 375
inquires whether the decremented count of the file metapage counter
equals the residual metapage count. If not, a copy metablock is
opened by the step 373. But if they are equal, that means that
there is not a metablock of data remaining in the current file to
be copied. Therefore, in a step 377, an open common metablock is
identified for copying residual data into it. The processing of
steps 363 and beyond then copies the remaining metapages of the
current data group of the current file are copied into a common
metablock. This path is not taken when it is an open file that is
being garbage collected, as determined by the step 371, since other
data groups could be written to the current file that would fill
out another metablock. After the open file is closed, it is placed
in a queue for garbage collection, at which time any residual data
are copied to a common metablock by the step 377.
[0243] An inquiry 379 of FIG. 31 determines whether the copy
metablock is now full, after writing this additional metapage of
data. If so, the copy metablock is closed in a step 381 and the FIT
is updated by a step 383 to reflect that fact. If the copy
metablock is not full, or after step 383, an inquiry 385 determines
whether all the data in the metablock being garbage collected have
now been rendered obsolete. If so, the metablock is erased, in a
step 387. If not, or after the metablock erasure, an inquiry 389
asks whether the foreground mode was set by the step 325 of FIG. 28
during a programming operation. If not, a step 391 of FIG. 31
determines whether a host command is pending. If so, and if the
garbage collection in progress is not a priority, as determined by
a step 392, the garbage collection is suspended by updating the FIT
in a step 393 and then returning to the step 253 of FIG. 26.
However, If a host command is not pending, as determined by the
step 391, or if there is a pending host command and the garbage
collection in progress is not a priority, as determined by the step
392, the processing returns to the inquiry 359 to copy the next
metapage of data from the current file metablock into the copy
metablock.
[0244] If the inquiry 389 of FIG. 31 determines that the foreground
mode has been set, a next step 395 increments the copy metablock
metapage counter that was reset by the step 327 of FIG. 28. If a
step 397 determines that a preset number N4 of metapages have been
copied in succession, then a next step 399 resets the foreground
mode and the FIT is updated by the step 393. But if there are more
metapages to be copied before the number N4 is reached, the process
continues with the step 359 unless the inquiry 391 determines that
there is a host command pending. If there is a host command
pending, the copy operation is interrupted and the process returned
to step 253 of FIG. 26.
[0245] Returning to the inquiry 359, if the data group length count
is zero, that means that copying of one data group from the file
metablock to the copy metablock has been completed. In this
circumstance, a step 401 updates the FIT to reflect this condition.
Next, it is determined in a step 403, by reference to the data
group number counter, whether the current file being garbage
collected includes another data group. If not, garbage collection
of the file has been completed. The process then returns to the
step 356 to garbage collect the next file or metablock in order
from a queue. It will not return to the step 355 since completion
of garbage collection of the current file means that there could no
longer be garbage collection of a file in progress and which could
be resumed.
[0246] If from the step 403 it is known that there is at least one
more data group of the current file to be copied, it is then
determined, in a step 405 whether the metablock containing this
next data group also includes other data that is obsolete. As
described earlier, metablocks containing obsolete data of a file
are garbage collected but those which do not contain obsolete data
of a file are preferably not garbage collected unless the metablock
is a common metablock or an incomplete metablock containing
residual data of the file. Therefore, if it is determined by the
step 405 that there are obsolete data in the metablock of the next
data group, the data group number counter is updated to the number
of the next in order data group, by a step 407, and the data group
length counter is initialized in a step 409 with the amount of data
in the new data group. The processing then proceeds to the step 361
to copy data of the first metapage from the new data group of the
file to the next erased metapage in a copy metablock, in the manner
previously described.
[0247] But even if it is determined by the step 405 that there are
no obsolete data in the metablock in which the next data group
exists, garbage collection of the data in that metablock can still
occur if either (1) the next data group is in a common metablock or
(2) is in an incomplete metablock and the file is a closed file. An
inquiry 411 is first made as to whether the current data group is
in a common metablock or not. If it is, then a next step 413 adds
the common metablock to the common metablock garbage collection
queue for later garbage collection, but the processing proceeds
with the steps 407 and 409 to update the data group number counter
and the data group length counter to that of the next metablock for
copying by the steps 361 et seq. However, if the current data group
is not in a common metablock, a step 415 inquires whether the
current data group is in an incomplete metablock. That is, the step
415 determines whether the current data group exists in a metablock
that still has at least a minimum amount of erased capacity (such
as the data group F3,D3 of FIG. 15). If not, the current data group
is not copied but rather the processing returns to the step 403 to
deal with the next data group of the file, if one exists. But if
the current data group is in an incomplete metablock, a next
inquiry 417 asks whether the incomplete metablock contains data of
an open file. If so, copying of the current data group is also
skipped by returning to the inquiry 403 to proceed with any further
data group of the file. But if the metablock in which the next data
group exists does not contain data of an open file, then the steps
407, 409, 361, et seq., are carried out on that next data
group.
[0248] Returning to the step 356 of FIG. 31, garbage collection may
take place even if it is there determined that no file or metablock
are present in a garbage collection queue. If there are not, rather
than ending the process, a step 421 checks whether there are more
than N2 erased metablocks in the erased metablock pool. If so, then
the garbage collection process ends and returns to the step 253 of
FIG. 26. But if there are not more than N2 erased metablocks in the
system, then the opportunity is taken to perform garbage collection
from the step 421. Such garbage collection can occur, as indicated
by a step 423, on an open file, if one exists. Since nothing is in
a queue, there is likely no other potential source of more erased
metablocks when the number of erased metablocks in the pool is N2
or less. If there is more than one open file, one of them is
selected in a step 425. The processing then proceeds for the opened
file through the steps 370, 372 and 359, as previously
described.
[0249] When it is determined by the inquiry 362 of FIG. 31 that the
current garbage collection operation is being carried out with data
groups in a common metablock, the garbage collection 364 is
somewhat different from that described above for other metablocks.
The flowchart of FIG. 32 outlines garbage collection of a common
metablock. A step 431 asks whether a garbage collection is in
progress and thus being resumed. If so, the data group length
counter retains the last value when the garbage collection was
suspended. If not, the data group number counter is first
initialized in a step 435 by reference to the FIT for the first
data group that is the subject of the processing and then the
length of this data group is then determined by the step 433.
[0250] When the data group length counter is determined by the step
433 to be greater than zero, copying of the current data group
commences. By a step 437, a metapage of the current data group is
read from the common metablock and, by a step 439, programmed into
an open common metablock. In a step 441, the data group length
counter is then decremented by one metapage. If doing garbage
collection in the foreground, as indicated by the step 443, a step
445 causes the copy metablock metapage counter to be incremented by
one. This counter keeps track of the number of metapages of data
that have been copied in the current sequence. If this count is
determined by a step 447 to be in excess of the predetermined
number N4, the process then exits the foreground mode in a step
449. This is followed by the FIT being updated, in a step 451, and
the processing then returns to the step 253 of FIG. 26.
[0251] If the copy metablock metapage counter is determined by the
step 447 of FIG. 32 to have a value of N4 or less, a next step
inquires whether a host command is pending. If so, the FIT is
updated in the step 451 and the processing returned to the step 253
of FIG. 26, in order that the pending host command may be executed.
If not, the processing returns to the step 433 of FIG. 32 to
inquire whether another data metapage exists in the current data
group and, if so, to copy it from the obsolete common metablock to
the open common metablock, and so forth. Similarly, if, in the step
443, it is determined that the garbage collection is not being
performed in the foreground, the step 453 is executed next. Unless
a host command is pending, as determined by a step 453, the
processing then returns to the step 433 to potentially copy the
next metapage of data in the current data group, regardless of
whether the copy metablock metapage counter is in excess of N4,
since the step 447 is bypassed. The garbage collection in this case
is being performed in the background.
[0252] Returning to the step 433 of FIG. 32, if the data group
length count is not greater than zero, it is then known that all
the metapages of the current data group have been copied. The FIT
is then updated, in a step 455, and it is determined by a step 457
whether there is another data group in the common metablock being
garbage collected. If not, the obsolete common metablock is erased,
in a step 459, and the processing returns to the step 253 of FIG.
26. But if there is another data group in the common metablock, the
data group number counter is updated by a step 461. The FIT pointer
from the common metablock index is then read, in a step 463, and
the FIT entry defined by that pointer is read, in a step 465. If
the FIT entry matches the current data group, as determined by a
step 467, an inquiry 468 determines whether the current data group
is part of residual data that has already been identified. If so, a
step 469 causes the data group length counter to be initialized to
the length of the data group as read from the FIT.
[0253] But if in step 468 it is determined that the current data
group is not in existing residual data, new residual data of which
it is a part is then identified, in a step 470. In a step 471, open
common block with sufficient space to store all of the data of the
new residual data is then identified. This prevents splitting
residual data of a file between two or more different metablocks
when the residual data contain two or more data groups. This could
otherwise happen if the two or more data groups are copied
independently. In that case, the first data group could be copied
to a common block with enough erased space to store that data group
but without enough space to also store the second data group. The
second data group would then be copied to a different common block.
This would be undesirable since residual data of a file should not
be split between two different metablocks. If they are so split,
garbage collection of the file or of the common blocks would then
take more time.
[0254] The data group length counter for the current data group is
then initialized by the step 469. The first metapage of data of the
current data group is copied from the common metablock into the
identified open common metablock, by the steps 437 and 439.
However, if the FIT entry is determined by the step 467 not to
match the current data group, the processing returns to the step
457 to determine whether another data group exists in the common
metablock.
[0255] The garbage collection shown in the flowchart of FIG. 32
continues until all the valid data groups of the current common
metablock have been copied to a previously erased metablock, which
then becomes a new common metablock, or to one or more open common
metablocks. Data groups of different files may be copied to
different open common metablocks. The common metablock was
previously placed in a garbage collection queue because it
contained an obsolete data group. The complete transfer of all
metapages of data of each valid data group in the common metablock
takes one pass through the steps of FIG. 32 for each such metapage.
This copying may be interrupted every N4 metapages if the garbage
collection is being done in the foreground, or when a new host
command is received when operating in either the foreground or
background.
[0256] As an alternative to permanently presetting the numbers N3
and N4, they may be changed by the memory system controller in
response to data programming patterns of the host, in order to
sustain a uniform data programming speed.
Various States of Memory Blocks During Operation
[0257] The diagrm of FIG. 33 shows various individual states of the
memory blocks or metablocks of the system, and transitions between
those states, within a direct file storage memory system of the
type described above.
[0258] In the left-hand column, a block 501 is shown to be in an
erased state, within an erased block pool.
[0259] In the next column, blocks 503, 505 and 507 each contain
some valid data but also have erased capacity in which data from
the host may be written. The Write Block 503 is partially written
with valid data for a single file, and further data for the file
should be written to this block when supplied by the host. The Copy
Block 505 is partially written with valid data for a single file,
and further data for the file should be written to it when copied
during garbage collection of the file. The Open Common Block 507 is
partially written with valid data for two or more files, and a
residual data group for any file may be written to it during
garbage collection.
[0260] Blocks 509 and 511 of the next column are full of file data.
The File Block 509 is filled with valid data for a single file. The
Common Block 511 is filled with valid data for two or more
files.
[0261] The next to the right hand column of FIG. 33 includes blocks
513, 515, 517, 519 and 521 that each contain some obsolete data.
The Obsolete File Block 513 is filled with any combination of valid
data and obsolete data for a single file. The Obsolete Write Block
515 is partially written with any combination of valid data and
obsolete data for a single file, and further data for the file
should be written to it when supplied by the host. The Obsolete
Copy Block 517 is partially written with any combination of valid
data and obsolete data for a single file. The Obsolete Open Common
Block 519 is partially written with any combination of valid data
and obsolete data for two or more files. The Obsolete Common Block
521 is filled with any combination of valid data and obsolete data
for two or more files.
[0262] An obsolete block 523 in the right hand column of FIG. 33
contains only obsolete data.
[0263] Transitions of individual blocks among the block states
illustrated in FIG. 33 are also shown by lines labeled with small
letters. These transitions are as follows:
[0264] a--Erased Block 501 to Write Block 503: Data of a single
host file are written to an erased block.
[0265] b--Write Block 503 to Write Block 503: Data for a single
file from the host is written to a write block for that file.
[0266] c--Write Block 503 to File Block 509: Data for a single file
from the host is written to fill a write block for that file.
[0267] d--File Block 509 to Obsolete File Block 513: Part of the
data in a file block becomes obsolete as a result of an updated
version of the data being written by the host in a write block for
the file.
[0268] e--Obsolete File Block 513 to Obsolete Block 523: All valid
data in an obsolete file block becomes obsolete as a result of the
data being copied to another block during a garbage collection or
of the file being deleted by the host.
[0269] f--Write Block 503 to Obsolete Write Block 515: Part of the
data in a write block becomes obsolete as a result of an updated
version of the data being written by the host in the same write
block, or of the data being copied to another block during a
garbage collection.
[0270] g--Obsolete Write Block 515 to Obsolete Write Block 513:
Data for a single file from the host is written to an obsolete
write block for that file.
[0271] h--Obsolete Write Block 515 to Obsolete File Block 513: Data
for a single file from the host is written to fill an obsolete
write block for that file.
[0272] i--Obsolete Write Block 515 to Obsolete Block 523: All valid
data in an obsolete write block becomes obsolete as a result of the
data being copied to another block during a garbage collection or
of the file being deleted by the host.
[0273] j--Erased Block 501 to Copy Block 505: Data for a single
file is copied from another block to an erased block during a
garbage collection.
[0274] k--Write Block 503 to Copy Block 505: Data for a single file
is copied from another block to a write block for that file during
a garbage collection.
[0275] l--Copy Block 505 to Copy Block 505: Data for a single file
is copied from another block to a copy block for that file during a
garbage collection.
[0276] m--Copy Block 505 to File Block 509: Data for a single file
is copied from another block to fill a copy block for that file
during a garbage collection.
[0277] n--Copy Block 505 to Write Block 503: Data for a single file
from the host is written to a copy block for that file when the
file is reopened during a garbage collection.
[0278] o--Copy Block 505 to Obsolete Copy Block 517: Part or all of
the data in a copy block becomes obsolete as a result of an updated
version of the data being written by the host in a write block for
the file, or of the file being deleted by the host.
[0279] p--Obsolete Copy Block 517 to Obsolete Block 523: All valid
data in an obsolete copy block becomes obsolete as a result of the
data being copied to another block during a garbage collection or
of the file being deleted by the host.
[0280] q--Write Block 503 to Open Common Block 507: Residual data
for a file is written to a write block for a different closed file
during garbage collection.
[0281] r--Copy Block 505 to Open Common Block 507: Residual data
for a file is written to a copy block containing residual data for
a different file during garbage collection.
[0282] s--Open Common Block 507 to Open Common Block 507: Residual
data for a file is copied from a different block to an open common
block during garbage collection.
[0283] t--Open Common Block 507 to Obsolete Open Common Block 519:
Part or all of the data for one file in an open common block
becomes obsolete as a result of an updated version of the data
being written by the host in a write block for the file, of the
data being copied to another block during a garbage collection, or
of the file being deleted by the host.
[0284] u--Obsolete Open Common Block 519 to Obsolete Block 523: All
valid data in an obsolete open common block becomes obsolete s a
result of the data being copied to another block during a garbage
collection or of the file being deleted by the host.
[0285] v--Open Common Block 507 to Common Block 511: A residual
data group for a file is copied from another block to fill an open
common block for that file during a garbage collection.
[0286] w--Common Block 511 to Obsolete Common Block 521: Part or
all of the data for one file in a common block becomes obsolete as
a result of an updated version of the data being written by the
host in a write block for the file, of the data being copied to
another block during a garbage collection, or of the file being
deleted by the host.
[0287] x--Obsolete Common Block 521 to Obsolete Block 523: All
valid data in an obsolete common block becomes obsolete as a result
of the data being copied to another block during a garbage
collection or of the file being deleted by the host.
[0288] y--Write Block 503 to Obsolete Block 523: All valid data for
a single file in a write block becomes obsolete as a result of the
file being deleted by the host.
[0289] z--Copy Block 505 to Obsolete Block 523: All valid data in a
copy block becomes obsolete as a result of the data being copied to
another block during a garbage collection or of the file being
deleted by the host.
[0290] aa--File Block 509 to Obsolete Block 523: All data in a file
block becomes obsolete as a result of the file being deleted by the
host.
[0291] ab--Obsolete Block 523 to Erased Block 501: An obsolete
block is erased during a garbage collection.
CONCLUSION
[0292] Although the various aspects of the present invention have
been described with respect to exemplary embodiments thereof, it
will be understood that the present invention is entitled to
protection within the full scope of the appended claims.
* * * * *