CA1187615A - Disk format for secondary storage system - Google Patents

Disk format for secondary storage system

Info

Publication number
CA1187615A
CA1187615A CA000412883A CA412883A CA1187615A CA 1187615 A CA1187615 A CA 1187615A CA 000412883 A CA000412883 A CA 000412883A CA 412883 A CA412883 A CA 412883A CA 1187615 A CA1187615 A CA 1187615A
Authority
CA
Canada
Prior art keywords
sector
replacement
sectors
data
defective
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired
Application number
CA000412883A
Other languages
French (fr)
Inventor
Barry L. Rubinson
Richard F. Lary
Edward A. Gardner
Mark A. Parenti
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Digital Equipment Corp
Original Assignee
Digital Equipment Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Digital Equipment Corp filed Critical Digital Equipment Corp
Priority to CA000475224A priority Critical patent/CA1203618A/en
Priority to CA000475223A priority patent/CA1205193A/en
Application granted granted Critical
Publication of CA1187615A publication Critical patent/CA1187615A/en
Expired legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B20/00Signal processing not specific to the method of recording or reproducing; Circuits therefor
    • G11B20/10Digital recording or reproducing
    • G11B20/18Error detection or correction; Testing, e.g. of drop-outs
    • G11B20/1883Methods for assignment of alternate areas for defective areas
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B20/00Signal processing not specific to the method of recording or reproducing; Circuits therefor
    • G11B20/10Digital recording or reproducing
    • G11B20/12Formatting, e.g. arrangement of data block or words on the record carriers
    • G11B20/1217Formatting, e.g. arrangement of data block or words on the record carriers on discs
    • G11B20/1252Formatting, e.g. arrangement of data block or words on the record carriers on discs for discontinuous data, e.g. digital information signals, computer programme data
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B20/00Signal processing not specific to the method of recording or reproducing; Circuits therefor
    • G11B20/10Digital recording or reproducing
    • G11B20/18Error detection or correction; Testing, e.g. of drop-outs
    • G11B20/1803Error detection or correction; Testing, e.g. of drop-outs by redundancy in data representation
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B2220/00Record carriers by type
    • G11B2220/20Disc-shaped record carriers
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11BINFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
    • G11B2220/00Record carriers by type
    • G11B2220/20Disc-shaped record carriers
    • G11B2220/25Disc-shaped record carriers characterised in that the disc is based on a specific recording technology
    • G11B2220/2508Magnetic discs

Abstract

Abstract In a disk type mass storage facility for data processing systems, a disk format which improves handling of defective segments of medium and reduces access time.
The format has three layers. A first, physical layer comprises the bytes, sectors and collections of sectors, as well as error detection and correction codes. A
second, logical layer is used to address the physical layer and to collect together sectors to form a multiplicity of separately addressable spaces, with each space having a distinct functional utility. At a third, functional layer the use of data fields in each space is specified. This layer governs the handling of bad blocks if required, and the use of certain format information.
At the physical layer, bits are collected into sectors, sectors are collected into tracks, tracks are collected into groups, and groups are collected into cylinders. These are logically, rather than physically, defined collections. A group is set of tracks wherein individual tracks in a group can be selected in the inter-sector rotation time. At the logical layer, the disk is divided into three main areas, each separately addressable. The first main area is broken down into two sub-areas. The first sub-area is available to the host computing system for its use; the second sub-area is used for replacement of bad blocks. The second is for format information. The third main area is for diagnostics.
Access time latencies may be minimized by proper selection of logical addressing - i.e., by selecting the number of sectors/track, tracks/group and group/cylinder-to take advantage of access characteristics.
Handling of bad blocks is controlled by a hierarchically layered process. A portion of each disk, distributed across the medium, is reserved as spare sectors to replace defective sectors. After a bad sector is replaced, future attempts to access the bad sector are redirected (i.e., revectored) to the replacement sector.
For the simplest revectoring, the bad block is replaced by a replacement block in a known location. If that cannot be done, multiple copies of the replacement block's header are stored in the bad block's data field and the copies are compared to find the replacement address. If the comparison fails, or the header cannot be read, a back-up table is available to match the available replacement addresses with the original address which was replaced.
A special code is used to identify blocks wherein the medium is good but the contents of the block are logically corrupted.
Various drive parameters are stored in the disk and are supplied to its controller on command, permitting the controller to adapt thereto. These include the starting address for each of the various areas, or spaces, of the disk, for example, its transfer rate, logical addressing structure, etc.

Description

- 1 ~a~s Cross-Reference to Related Apr~ications This application relates to a data processing system~ other aspects of which are described in our Canadian application Serial No. 412,~26, titled [nterface Between a Pair of Processors, Such as Host and Peripheral-Controlling Processors In Data Processing Systems, filed October 4, 1982 and Serial No.
412,823, titled Storage Facility Employing Serial Communication Between Drive and Controller, filed October 5, 1982.
_eld of the Invention This invention relates to the field of data processing systems and, in particular, to the formatting of disk type mass storage facilities in such systems. This invention further relates primarily to such facilities which are fixed block, rather than variable block architecture.
Background and Summary of the Invention , Secondary storage subsystems, such as disk drives, are an important part of modern data processing systems. Such subsystems provide a large volume of memory for storing programs and data. In disk drives, rotating disks with magnetic recording material provide the actual storage medium.
A primary objective in the use of such secondary storage subsystems is to minimize the time required to read or write information at a specific address on a disk surface from a starting point at another address position.
The access time to move a read/write head to the desired target address is a function both of physical $:. qg:

~,, ,, :~

parameters of the disk drive (e.g., how fast the drive's electronic con~rol circuits can determine and supply appropriate signals to that actuator) and of the addressiny scheme employed (which will determine the physical spacing between starting and target addresses).
Another objective of such subsystems is to achieve high reliability in writing and reading data.
Unfortunately, the medium is not perfect. Portions of the oxide ~ur~ace of the medium may be manufactured defectively; other portions may degrade and wear out under c~nditions of long-term use. If information is written (i.e., recorded) on such areas, it cannot be stored or read (i.e., retrieved) reliably.
Error detection and correction techniques are, of course, part of the solution to this problem. However, error detection and correction may not be enough where the medium will not permit the recording o a sufEicient portion of a block so as to allow those techniques to be invoked successfully when the block is read. It is therefore important to avoid the use of portions of the medium which are found to be so bad that informati~n will be unrecoverable or where the information may degrade to an unrecoverable state. In the prior art~ several approaches or techniques have evolved for dealing with this problem.
A first technique simply invalidates an entire track when too much of it is bad. All of ~he information intended for that track is redirected to a substitute track. It will be readily apparent that this scheme may discard a lot of good medium with the bad. Further, only a limited number of substitute tracks can be made available without significantly detracting from the usable volume of medium.

~8'7~

A second technique, which is much less drastic, invalidates the bad sector and does not use bad blocks.
This, however, creates problems when transferring the contents of one disk surface to another disk surface, since it is statistically almost impossible to find the same bad blocks on two different surfaces. An additional disadvantage of this technique i5 that it causes holes in the logical addressing space.
A third technique is to provide on each track a limited amount of space which can be used to substitute for bad portions of sectors on thak track by skipping over the defective area and pushing the remainder of the sector further down the track. ThiS technique is helpful only up to the point where the defective area on a track does not exceed the reserved portions. It also causes sectors on different tracks to lose their alignment, causing problems in achieving real time head switching.
A fourth technique is to reserve i'n" sectors per track. Bad blocks are then either revectored (i.e., redirected) to one oE those sectors on that track, or all blocks subsequent to a bad block are "slid" down, without revectoring. This limits replacement to those sectors, per track however.
A fifth technique is to reserve some portion of the disk and to revector from the bad blocks to the reserved region ~hrough a table. This approach has the disadvantage of poor performance.
Since bad blocks can occur both during manufacturing and then subsequently during the use of the disk, it is important that bad block replacement be performed both initially, before the medium is first used to store host information, and later, when dynamic conditions give rise to appropriate circumstancesO Prior art techniques are not very good for both cases.

The present invention deals with this problem in a hierarchial, mul~i-level fashion. An evenly distributed portion of each disk is reserved as spare sectors for replacing defective sectors. After a bad sector has been replaced, future attempts to access the bad sector are redirected (i.e., re vectored) to the replacement sector.
Three levels of revectoring mechanism are illustrated;
they differ in the way that the address of a replacement block is determined. It is possible, optionally, to trade off performance against complexity by electing not to employ all of these mechanisms.
In the primary revectoring mechanism, the position of the replacement block is implied by the position of the bad block and the need to revector is indicated by a code in the header. Each track is provided with one or more replacement sectors. The implied primary replacement block for a bad block is the first replacement sector on its track. In the secondary revectoring mechanism, the need to revector is signalled ' 20 by a code in the header. 4~e location of the replacement block is arbitrary. To determine its address, multiple copies of the replacement block's header value are stored in the data ield of the bad block. The coPies are read and compared statistically to come up with'addres~ so indicated. Finally, there is a so-called tertiary revectoring mechanism used when the header copy comparison fails to yield a valid value or when the multiple copies of the replacement address in the secondary scheme do not meet the statistical matching requirement. For implementation of this mechanism, there are stored on the disk multiple copies of a table containing a list of each replacement block and the address of any bad block mapped to it, if any. This ~7~

table is searched to find the appropriate replacement address.
A uni~ue logical addressing scheme also is employed, collecting sectors according to a hierarchy of geometrical and access time considerations. This permits sectors to be addressed logically, rather than physically; they are sel~-defining in terms of physical locations, so as to optimize sector access ~ime latencies. This, combined with revectoring, provides a logically contiguous address space at all times i.e., one without holes.
A further aspect of this invention is that the disk is divided into different regions which comprise separate logical areas - one available to users, one for replacement of bad blocks, one for diagnostics, and one for recording certain information regarding disk formatting. Each i5 a logically self-consistent, but different, addressing space.
Initially, a disk is "inspected" for sectors which are bad when the disk is manufactured. These are replaced during the manufacturing process or at installation. Other sectors are replaced as they start to degrade in quality, but before they produce an error rate exceeding the capabilities of the error correcting code (~CC) which is employed. (~his ECC "threshold" is specified by the drive itself.) other sectors are replaced after they degrade and are not readable; this requires notification ~hat the data is corrupted.
Yet another feature of this invention is the use of a special code to distinguish sectors which contain logically corrupted information, but wherein the medium itself is usable. This special code is referred to as the "forced error" indicator; in ~he implementation described below, it is the one' 6 complemen~ of an error L87~

detecting code (EDC) genera~ed by ~he information in a sector's data field in accordance wi~h a preselected i~ algorithm~and appended to the data field of the sector.
When a sector is read, its EDC is computed and compared with the EDC recorded on the disk. If the comparison reveals that the EDC field is recorded as the one's complement of the computed EDC, the forced error indicator has been detected~ The host is thereby notified that the data is logically bad, but the medium is not known to be impaired. This indicator is use~ul, for example, during an offline volume copyl when the data in the block is found to be physically corrup~ed and uncorrectable, but must be copied to a physically good sector on another volume. In order ~o allow hosts which access the copy to know that the data there is corrupted and unreliable, the forced error indicator is set in that sector. The nex~ time information is written to this sector~ the forced error indicator will be cleared, since the medium itself is good and only the information previously written to the medium was bad.
Use of the forced error indicator follows ~hree rules. first, a read operation from a block where the forced error indicator is set must alwa~s fail. Second, a wri~e operation to such a block must clear the forced error indicator~ Third, a read operation must produce a unique error code so as to differentiate the detection of a forced error from any other read error.
It should be appreciated that this forced error indicator is not part of the data bytes transferred; it is control information generated when the sector is written.
The contents of certain portions of the disk which are not protected by replacement are protected by virtue of being written in multiple locations, to store multiple ~$

copies of the same information. If a sufficient number of copies, or portions of copies, are recorded unimpaired, the recorded in-formation can be retrieved despite the corruption of one or more copies.
In summary, according to a first aspect of the present invention, there is provided in a secondary storage subsystem for a data processing system, wherein data is recorded on a mass stor-age medium and the smallest addressable unit of the medium is a sector, each sector including a ~Jeader field for recording address information and a data field for recording data to be associated with and stored at such address, the improvement comprising: means for writing in each sector a prede~ermined code, termed the forced error indicator r when the data being recorded in the data field is known to be logically corrupted and the medium underlying the sector is not known to be defective; and means for providing a signal, termed a forced error signal responsive to detea~ion of the forced error indicator during a read operation.
According to a second aspect of the present invention, there is provided in a secondary storage subsystem for a data pro-cessing system, wherein data is recorded on a mass storage mediumand the smallest addressable unit of the medium is a sector, a method of replacing a defective sector with a substitute sector, such that information to be written to or read from a defective sector is written to and then read from the substitute sector in-stead, once the defective sector is identified as unreliable, such method comprising the steps of: A. reservlng a portion of the medium ~.o be used as spare sectors for replacing defective sectors, at least one spare sect.or being provided within each set of sec-tors occupying contiguous logical locations on the medium; B. re-placing the first defective sector i.n said set of sectors with the first one of said spare sectors, termed a primary replacement sector; C. indicating such replacement by writing a first predeter-m:ined code in the defective sector; and D. when writing data to or reading data from a sector, detec-ting the Eirst predetermined code and, in response to detecting said code, revectoring the writ-ing or readlng operation to said first one of the spare sectors for the involved set of sectors where the code was detected.
According to a third aspect of the present invention, there is provided in a secondary storage subsystem for a data pro-cessing system, wherein data is recorded on a mass storage medium and the smallest addressable unit of the medium is a sector, each sectol comprising a header field and a data field, the address of the sector normally being written in the header field and the information to be stored n the sector being written in the data field, a method of replacing a defective sector with a substitute sector, such tha-t information to be written to or read from a ae--fective sector with a substitute sector, such that informationto be written or read from a defective sector is written to and then read from the substitute sector instead, once the defective sector is identified as unreliable, such method comprising the steps of: A. re~erving a portion of the medium to be used as spare sectors for replacing defective sectors; B. selecting for a defec-tive sector a replacement sector, from among said spare sectors;
C. in the header field of such defective sector, writing a pre-` -7a-determined code indicating that said sector has been replaced by a secondary replacement sector; D. in -the data field of a defective sector, writing a predetermined multiple number of copies of the physical address of -the selected replacement sector -therefor, ter-med a secondary replacement sector; E. on reading the header field of the defective sector, checking for said second code; F. respon-slve to detecting said second code, obtaining -the physical address of the secondary replacement sec-tor by reading said multiple copies and comparing -them statistically to arrive at the recorded value of said address; and G. revectoring the writing or reading opera-tion intended for the defective sector to said secondary replace-ment sector.
According to a fourth aspect of the present invention, there is provided in a secondary storage subsystem for a data pro-cessing system, wherein data is recorded on a mass storage medium and the smallest addressable unit of the medium is a sector, a method of replacing a defective sector with a substitute sector, such that information to be written to or read :Erom a defective sector is written to and then read from the substitute sector in-stead, once the defective sector is identified as unreliable, suchmethod comprising the steps of: A. reserving a portion of the medium to be used as spare sectors for replacing defective sec-tors; B.
se ecting for a defective sector a replacement sector from among said spare sectors; C. providing on -the medium multiple copies of a table containing a list of each spare sector and the address of the defective sector replaced by it, if any; D. responsi.ve to detec-ting a defective sector, searching a copy of said table to -7b-I il7~i ~ ~ .

find the address of the replacement sector therefor; and E. revec-toring the writing or reading operation intended for the de:Eective sector to said replacement sector.
The invention will now be described in greater detail w.ith reference to the accompa.nying drawings, in which:
Figure lA is a generalized block diagram of a data pro-cessing system in which the present invention may be utili~ed;
Figure lB is a diagrammatic illustration of a disk sur-face in the disk drive of Figure lA, according to the invention;
Figure lC is a diagrammatic illustration of the logical spaces provided on a disk, according to the present invention;
Figure 2 is an exemplary procedure for reading informa-tion protected by the multi-copy mechanism disclosed herein;
Figure 3 is an exemplary procedure for writing informa-tion onto a disk according to the multi-copy protection mechanism;
Figures ~A-~D are a flow chart of the bad block replace-ment procedure of the present invention;
Figure 5 is a diagrammatic illustration of the format of a replacement block descriptor;
Figure 6 is a diagrammatic illustration of the struc-ture of the replacement and caching table (RCT) of the present invention;
Figures 7A and 7B are a diagrammatic illustration of the contents of sector zero of the RCT of Figure 6;
Figure 8 is a diagrammatic illustration of the RCT
search algorithm of the present invention;
Figures 9A-9C, together, are a listing of an exemplary procedure for implementing the RCT search algorithm of Figure 8;

-7c-~,.

Figure 10 is a listing of an exempla~y procedure for implementing the RCT hash algorithm described herein;
Figure 11 is a diagrammatic illustration of the sector :Eo:rmat employed by the present invention;
Figure 12 is a diagrammatic illustration of one-fourth of the sector header of the sector format of Figure 11;
Figure 13 is an exemplar~ listing of an error detecting ~ode usable in the present invention;
Figure 1~ is a diagrammatic illustration of a response from drive to controller, providing certain dri~e characteristics;
Figures 15A and 15B, together~ are a diagrammatic illus-tration of a response providing certain other characteristics for a designated drive subunit;
Figure 16 is a diagrammatic illustration of the first two and the last tracks of the LBN/RBN space of:a.dEive;
Figure 17 is a diagrammatic illustration of the first two and the last tracks of the XBN space of a drive;
Figure 18 is a diagrammatic illustration of the first two and the last tracks of the DBN space of a drive;
Figure 19 is a listing of a procedure for determ~ning a logical block's sector address;
Figure 20 is a listing of a procedure for determining a replacement block's sector address;
Figure 21 is a listing of a procedure for determining an external block's sector address;
Figure 22 is a listing of a procedure for determining a diagnostic block's sector address;

- ` ~

Figure 23 is a listing of an exemplary procedure for reading 128 copies of an RBN's header to determine reliably the correct value of that header;
Figure 24 is a diagrammatic illustration of the format of a copy of the FCT (format control table) in the XBN space of th~ present invention; and Flgure 25 is a diagrammatic illustration of sector zero of each FCT copy.
~ eferred Embodiment A generalized block diagram of a data processing system which can utilize the present inventlon is shown in Figure 1.
That system includes a host computer 1 (having a central proces-sor, main memory and input/output devices, not shown) and a secon-dary storage subsystem 2~ The secondary storage subsystem, in turn, includes a disk drive 3 and a controller 4 for that disk drive. The controller 4 typically contains one or more processors of its own and provides proper signals for actuating the drive to read or write information in accordance with instructions from the host processor.
The present invention utilizes a unique disk format to improve defect handling and reduce access time. It ' ~

~ 76~
- , g ., s.

further serves to protect certain selected areas of a disk by storing multiple copies, and facilitates adaptation to peripheral device timing characteristicSO
This format comprises an architecture composed of three layers. First, there is a physical layer, comprising the actual bytes, sectors, and collec~ions of sectors on the disk. This layer also includes an error detection and correction mechanism. Second, there is a loyical layer, which is the level at which the physical layer is addressed and grouped into spaces, with particular uses assigned to those spaces. Third, there is a functional layer, which is the level at which the use of data fields in each space is described. This layer includes the handling of bad blocks, if required, and other format usage information~
What follows generally is a detailed description of each of those layers. Before proceeding, however, it will prove useful to define certain terminology used herein.
A "sector" is the smallest unit by which data is addressed physically on a disk surface. Each sector occupies a specific physical position relative to an index location on a disk, and has the property that it is available ior reading or writing once per disk rotation.
Sectors are grouped together hierarchically for addressing purposes. First7 a disk surface is divided into one or mo~e "cylinders." In turn, cylinders are divided into "groups;" and yroups are divided into "tracks."
A ~Itrack" is a logical entity comprising a set of sectors occupying contiguous logical disk locations.
A "group" is, in turn, a logical entity representing a set of tracks such that individual tracks in a group can be selected within the inter-sector rotation time.

10 ~ 37~

Tracks in the same group are "aligned" such that the same physical sector address is available simultaneously for reading or writing on all tracks in the group.
A "cylinder" is a logical entity representing a collection of groups which can be selected via an operation with latencies less than the minimum "seek"
time. Cylinders have the addi~ional property that the selection of a new cylinder requires the longest average head-positioning operation. Groups within a cylinder are offset such that spiraling between adjacent graps is accomplished without the loss of a full revolution of the disk~
~ he terms track, group and cylinder simply relate collections of sectors to each other as a function of access time considerations. They are independent of physical organization or construction of the device.
The "sector number" portion of a sector address is always the low order portion. The ~track number" portion of a specific sector address is always the middle portion of that address between the group and sector portions.
The "group number" por~ion oE a specific sector address is always a middle portion of that address between cylinder and track. The "cylinder number" portion of a specific sector address is always the highest order portion of that address.
A ~bad block" is a sector containing a defect which either (1) causes an error exceeding the correction capability of the error correction scheme used by the subsystem, or (2) exceeds a drive-specified error threshold beyond which the continued integrity of data in the sector cannot be guaranteed. A bad block may also be a sector which is defective to such an extent that, while data integrity is still assured, the overhead imposed by ~711~$

the required error correction procedure i5 greater than will be tolerated~
"Bad block replacement" is the designation of a spare sector (i.e., replacement block) to substitute for a bad block.
"Bad block revectoring" denotes the act of transferring a read or write operation to the replacement block associated with a bad sector upon access thereto.
A "physical block number" (PBN) is a number which identifies a sector's physical position within the set of sectors on a mass storage device.
A ~logical block number" (LBN) is a number identifying a sector's relative position within the set of sectors directly accessible to the host. These are used for host data storage and revector control information A "replacement block number" (RBN) is a number which identifies a sector's relative position within the set of sectors used as replacements for bad sectors.
A "primary replacement block" is a replacement block with a designated RBN on a track which has been allocated to replace a logical block on the same track~
A "secondary replacement block" is a replacement block which is not a primary replacement block~ It is either not the replacement block with the designated RBN
of the primary replacement block on a track or is allocated to replace a logical block located on another track.
An "external block number" (XBN) is a number which identifies a sector's relative position within the set of sectors in the external format area~
A "diagnostic block number" (DB~) is a number which identifies a sector's relative pOSitiQn within the set of sectors in the diagnostic area.

6~i A "host" is a central processing unit serviced by a secondary storage subsystem~
A drive may be in any of several states relative to the controller. When in the 'drive-offline" state, the drive is not operational and may not communicate with the controller via the drive control protocol. Conversely, a "drive-online" drive is dedicated to the exclusive use of a particular controller and is not available to any alternate controller~
In the "drive-unavailable" state, the drive is operating, is visible to~ and may at times communicate with the controller; but the controller may not fully utilize the drive because the drive is "drive-online" to another controller.
A drive which is "drive-available" is one which is visible to, capable of communicating with, and capable of becoming (but is no~ currently) "drive-online" to any specific controller.
It will also be helpful to define briefly a number of generic symbols which will be used thorughout this description and in the accompanying drawings:
S~mbol Meanin~

C - Starting cylinder L a Starting LBN
R - Starting RBN
X = Starting XBN
D = Starting DBN
1 = L~N's per track r = RBN's per track s = sectors per track `~ ~

t = tracks per group g = groups per cylinder o = Group offset H = LBN's in host area Lc = LBN space size in cylinders Xc = XBN space size in cylinders Dc = DBN space size in cylinders Rct = Size of one copy of replacement and caching table, in LBN~s n = Number of copies of tables in non-revectored spaces Rs = Lc*g*t*r the number of replacement blocks Rctpad = Size of per RCT copy track/cylinder alignment pad Rsp = Replacement and caching area size (n*Rct-~Rctpad*(n-l)) The Physical Layer The sector is the basic addressable unit of the disk. A
disk 5, as shown in Figure lB, is a circular platter having a coating of a ferromagne-tic material on a rigid substrate. For reading information from the disk or writing informatlon to the disk, a transducer head 6 is positioned by an actuator 7 over one of a number of concentric bands whose center lines are indicated by the numerals 8. Each "sector", such as a sector 9, is an arcuate segment of such a band, of finite extent. Every sector is written in a predetermined format which includes a header, data bytes, error detecting code and error correcting code. Each , ~

header is, in turn, a 32-bit quantity that contains the logical address of the sector, There~are four copies of the header in every sector. The data bytes are application-specific information recorded on the disk by host and su~system input/output opera-tions. By convention there are either 512 or 576 bytes of data in every sector, when standard Eormats are employed. (For example, the assignee's PDP-ll and VAX-ll computer systems use a 512 byte format~ while its PDP-lO and DECSYSTE~-20 computer systems use a 576 byte format; PDP, VA~ and DECSYSTEM are trademarks of Digital Equipment Corp., Maynard, Massachusetts, U.S.A.) Sector layout is described in greater detail below.
"Tracks!', "groups'~ and "cylinders" are collections of sectors grouped into a hierarchy of categories according to access time latencies. Access time to any sector on a -track is a linear function of the distance of that sector from the current sector which is under the read/write head, if on the same track. The first sector on the track immediately follows the last sector with respect to access time considerations. These properties con-strain a track into the logical (but not necessarily physical) structure of a ring. Similarly, a group is described as a col-lection of tracks wherein the time required to switch from a sec-tor at a given angular position on one track to a sector at the next sequential angular position on any other track is less than or equal to the time required to traverse the gap between adjacent sectors on the same track, during normal rotation.
Customarily, in a disk drive a single head-posit.ioning actuator will be used to position multiple read/write heads which are separated from each other by a fixed distance. When instruc~
ted to read or write, a controller determines which of the heads services the addressed portion of the disk and uses that head to perEorm the operation.
Traversing the distance between two logically adjacent tracks in adjacent groups may involve a head switching time which exceeds the gap time. Therefore, according to this invention, in order to not lose a revolution of the dlsk while the switching occurs, the first sector (i.e., the sector with the lowest PBN) on all tracks on the next higher-numbered group is offset angularly from the first sector on all tracks of the next lower-numbered group on the same cylinder by a number of sectors sufficient to compensate for the head switching time.
~ or cylinders having more than one group, the sector with the lowest physical block number on a track in the second group of the cylinder is offset from the index -l~a~

mark on that track by a number of sectors representiny at least the maximum switching time between the two groups (modulo the rotation time). The third group on the cylinder would be offset by twice this value, etc.
Three examples will illustrate this logical addressing structure and definition:
Assume first that a drive has four disks and seven physical oxide surfaces used for data storage. Each data recording surface has two read/write heads associated therewith; switching between one physical head and another head on the same or any other oxide surface can be accomplished during the intersector time. Other than selecting one of the fourteen heads, however, there is nothing else that can be accessed without moving the heads via a cylinder switching (seek)O ~s a result, the drive has a logical geometry of "14 tracks~ one group, 560 cylinders."
Assume next the same physical configuration as above, implemented using a servo technology which requires a settling time larger than the intersector time when a head is selected. Like the above drive, this drive would have seven physical oxide surfaces used for data storage and two heads on each surface. Switching between one head and any other head, however, would result in a head settling time greater than any intersector time, regardless of which oxide surface the head resides on. A switching of heads on such a drive would thus be accompllshed via a group switching operation (group select)~ Other then selecting one of the fourteen data heads, however, there is nothing else that could be accessed on the current cylinder without moving the heads via a seek operation. The logical geometry of such a drive, thereEore, is "1 track, 14 groups, 560 cylinders."

' '` . ' .

16 ~.1~7~

Third, assume semiconductor technology which provides thin-film heads capable of accommodating several read/write gaps on a single head. If the above mentioned hypothetical drive were fitted with ~uch heads, each physical arm would have multiple gaps located on it, each selectable in the intersector time. For discussion, assume eight such gaps per arm. Inter-head changes would be accomplished via group select operations. Such a device would thus have a logical geometry of "8 tracks, 14 groups, 560 cylinders."
All three of the above logical geometries, of course, were derived from the same physical geometry.
It is possible by changing the logical organization of a disk, to change its access time characteristics.
Xndeed, proper redefinition of the logical addressing structure has been found to yield a possible average access time reduction of 5 to 6 percent, which is due to minimization of rotational access latencies. It is difficult to generalize as to a technique for selecting the best logical addressing arrangement for a particular drive, since ~here are a large number of physical parameters involved and some relate to host characteristics. So far, trial and error based on knowledgeable guesswork has proven successful. That is, a drive is exercised with a large number of read and write operations and its average access time are determined. A change in the logical structure is made (e.g., the number of groups is altered), the disk is reformatted and the drive is again exercised. This process is repeated as desired. The result is a logical format "tuned" to the physical characteristics of the host-controller-drive combination.
The Logi_al Layer Attention is now directed to the logical layer of the format. At t~is layer, the disk is divided into four address spaces of interest, as illustrated in ~
Two of these address spaces are in the set of sectors having an effect on the host; the other two are invisible to the host. (Specific implementations may have additional address spaces which are not visible to the host). The first address space 12 contains the set of LBN's which are visible to the host. This LBN space is subdivided into t~o areas~ the host applications area 12A and the Replacement and Caching Table (RCT) area 12B.
For a given sector size, the host applications area 12A
is of constant size with respect to the number of usable blocks; that is, it is "bad block free." The RCT area 12B is not bad block free; it is protected by a multi-copy error handling mechanism described belowO
The RCT address space contains a list of RBN's which are used to replace LBN's that have become unusable on devices susceptible to bad blocks~ These RBN's comprise a second logical space, 14, and are the last r sectors of each track (where l'r" is a drive-specific parameter) in host applications area 12A. RBN space 14 is outside the L8N space 12 presented by the controller to the host so RBN's are invisible to the host (except for the performance implications they may have on allocation policies and ~he size of the RCT area 12B of LBN space 12). A replacement block number is converted to a specific physical device location by a series of transf~rmations per~ormed by the subsystem using parameters supplied by the storage device to the subsystem. These conversions are subsystem implementation-dependent and are of no interest to the host.

~ 18 ~7~

The area 16 contains the XBN's which provide formatting information. It is inaccessible to the host and always is written in a predefined format (e.g., 512 bytes/sector) even if other areas (eOg., LBN space 12) may be written in other formats. The contents of area 16 are multiple copies of format control information and media error lists to use when formatting the disk in such alternative formats as may be available (e.g., 512 or 576 bytes/sector). The media error lists comprise a Format Control Table (FCT), which is a table containing information regarding blocks found to be bad at the time of manufacture.
Area 18 contains the diagnostic cylinders (DBNIs).
This area is inaccessible to the host and is utilized solely by diagnostics executing out of ~he mass storage subsystem.
The mass storage subsystem is responsible for utilizing the logical blocks and replacement blocks in a fashion presenting the host with a logically contiguous set of blocks numbered from 0 to ~-1 for each unit/ where H is the block capacity of the host applications area of the unit, as seen from the host. Allocated replacement blocks belong to one of two performance classifications-(1) primary replacement blocks and (2) secondary replacement blocks. Primary replacement blocks are those allocated in the simplest (and usually fastest) manner, such that a negligible amount of time (i.e~, at most one rotation) is used to access them during revectoring.
Secondary replacement blocks are those that are determined in a more complex fashion and normally require greater than one rotation time to access them during revectoring (i.e., a group selection or seek).
The Functional Lay~

~q~

The functional layer comprises bad block handling. Two bad block handling mechanisms are used on the media. These are:
(1) the use of multi-copy structures and (2) replacement and revectoring. The former protects from failure by recording mul-tiple copies of important fields; the letter is a mechanism that replaces bad sectors with spare sectors which are reserved for that purpose. These mechanisms are used in different areas and have two fundamentally different effects.
Multi-copy structures are allocated in the RCT and FCT
area to provlde protection for critical data structures. Replace-ment/revectoring is used in the host applications area to elimin-ate holes in the address space.
Multi-copy structures are allocated in areas where re-placement is either not possible or not desirable. An example of a multi-copy structure is the RCT described herein. In reading or writing a multi-copy structure, only one block of the logical structure may be accessed at a time. Each copy must be accessed sequentially, in ascending order. Error recovery and correction must be enabled for both read and write operations. The number of copies allocated is a device characteristic and must be chosen to ensure achievement of the error rate goals of the system, A
typical value for n will be four. Each copy must be positioned on the physical medium such that a single failure will not inval-idate all copies.
The Multi-Copy Read Algorithm _ .~ ~ .,_. . _.. ._ ._.
Figure 2 is a listing of an exemplary computer program illustrating the method used to read sectors protected by the multi-copy protection mechanism. It provides a procedure which detects errors and attempts to read the ne~t copy in sequence.
For this method to operate properly, error correction and recovery must be available and in ~se. In the Figure, the variable "target"
represents the address of the sector being read in the first copy;
"copy-size"

19a-is the size of a copy of the information being protected, including any pad, "n" is the number o~ copies; "next" is the next copy to examine; and "data-blk" is the block into which one sector's worth of data is read.
The Multi-Co~y Write Algorithm Fig. 3 illustrates the corresponding method utilized to write sectors protected by the multi-copy protection mechanism. As with the multi-copy read algorithm, copies are accessed sequentially and error reco~ery must be enabled. In that Figure, the variables have the following meanings: "target" refers to the address of the sector in the first copy where the information is to be written;
"copy-size" is the size of a copy of the information being protected, including any pad; "n" is the number of copies; "next" is the next copy to write; "err-count" is the number of write failures; and "data-blk" is the block of data (one sector) to write Replacement/Revectoring Protection Replacement is used under three circumstances to substitute spare sectors (RBN' s) for sectors in the host application area (LBN's): (1) to fill holes in the logical address space left by bad blocks; (2) to reduce the risk of failure due to progressive deterioration of sectors; and (3) to imyrove performance in those implementations wherein the correction mechanism (if any) requires more time ~han the revectoring mechanism. The secondary storage subsystem determines the occurrence of the above-listed cases and initiates replacement. This is done either by notiying the host to begin host-initiated replacement or by performing subsystem-initiated replacement.
Notification to begin host initiated replacement is done by a predetermined host protocol message packet. In the case of an unrecoverable error, the packet contains _~ 21 both the failure notification and the bad block notification; in the other cases, it contains only a bad block notification~ If bad block replacement is host-initiated, the bad block may be replaced immediately (termed dynamic replacement) or when the file or data structure ls re-allocated ~termed static replacement).
Dynamic replacement is made possible by the "forced error indicator", by writing the replacement block with the "forced error" modifierO If subsystem ~ initiated replacement is used, it is dynamic.
After a sector has been replaced, revectoring will occur upon each host access to the replaced LBN.
Replacement Strategy In explaining bad sector replacement, certain terminology must be understood. First, if an error is said to be "recoverable" and "correctable", that implies the associated operation can be performed successfully;
an operation fails if and only if an "unrecoverable"
error occurs. Second, the disk going "offline1' or "available" in the middle of bad block replacement is treated as an unrecoverable error; the disk must not be brought back online with the bad block replacement algorithm resuming where it left off. Instead, when the host or controller next brings the unit online, it must act on the data stored in RCT sector zero exactly as it would any other time a disk is brought online.
Bad sectors which are discovered during the media formatting process are replaced at that time. ~ad blocks which are caused by wear are replaced according to the procedure detailed below in Figs. 4A-4D. A two phase replacement scheme is utilized. First (i.e., phase 1), a block is determined to be bad. Second (ie., phase 2), the bad block is replaced.

When a specifled logical block is found to be bad during an attempt to read that block, the bad block replacement process is notified of that fact. ~Step 110). Alternatively, the bad block replacement process may be notified that a failure or loss oE context occurred in the middle of bad bloc]c replacement; this is detected while bringing the unit online. Step 112. Either notification includes the bad block's LBN, the previous data con-tents of the bad block, and an indication as to whether or not the data is valid (i.e., whether the original reading of the bad block succeeded). In the case of a failure or loss of context during phase 2, the notification also includes the new RBN with which to replace the bad block, whether or not the bad block had previously been replaced, and (if it had been previously replaced) the old RBN that currently replaces the bad block.
Second, the bad block replacement process locks out all access to the bad block and all update or write access to the unit's RCT. Step 114. This is a "soft" lock - i.e., one which is im-plicitly released if the host or subsystem (whichever is perfor-ming the replacement) loses context. Optionally, since bad block replacement frequency should be low, it may be acceptable to lock out all access to the entire unit, rather than just -the bad block and RCT.
Third, step 116, a determination is made as to what type of loss of context or failure is involved. If it is one which occurred in phase 2, then control is relinquished to step 1~4. If the failure or loss of context occurred during phase 1, though, the process branches to step 132. If there was no loss o:E context, ~22-~ .

the process continues wi h step 118.
Fourth (step 118), a sector-sized buffer is cleared and the current contents of the bad block are written -22a-23 ~ IB76~5 into that buffer. (The buffer is initially cleared to account for the rare case wherein no data can be transferred.) A read operation is the performed, step 120, with error recovery and error correction capabilities enabled and evaluated as to success, step 122. The saved data is considered valid if the read succeeded, and invalid if it did not.
Next, step 124, the data obtained when the bad block was read during step 120 is recorded in the second sector (i.e., sector 1) of each RCT copy and evaluated for success. Step 126. If the data cannot be successfully recorded in the RCT, the error is reported to the error log, step 128, and process con~rol is transferred to step 174.
Sixth, in step 130, the bad block's LBN is recorded, with an indication o~ whether or not the saved data is valid, together with the fact that the process is in phase l; this information is put in sector zero of each RCT copy. This, of course, requires reading sector zero, modifying the appropriate flag (Pl), then writing the updated sector zero in order to preserve other information stored in that sector. If RCT sector zero cannot be successfully read, the error is reported to the error log and control is transferred to step 174. If sector zero cannot be successfully written, the error is reported to the error log and sequence control is transferred to step 170 Seventh, step 132, test patterns are written to and read from the suspected bad block to determine whether or not it actually is a bad block. If the test patterns fail, control is transferred to step 136 (and the block is presumed bad). If the test patterns succeed, then the - process continues with step 13~, under the assumption that the block may be good. The test patterns fail if ~87~

either (1) the block is reported as a bad block for the second time or ~2~ the test patterns cannot be written and read back correctly.
In step 1~4, the saved data is written back to the bad block using a write-compare operation. The write-compare is performed with the "forced error" flag if and only if the saved data is invalid. IE the write-compare both succeeds and the block is no longer reported as a bad block, then the original problem was a transient one and the sequence resumes at step 156. The write-compare succeeds if no error is detected and the saved da~a ~s valid, or if only a forced error is detected and the saved data is invalid.
The next step, step 136, is to scan the RCT and determine the new RBN to use to replace the bad block, whether or not the bad block has been previously replaced, and the bad block's old RBN if it has previously been replaced. The RCT is not updated at this time. If the RCT scan fails, the error is reported to the error log; step 138, and control jumps to step 166.
If there is no jump, step 140 is performed next. There, the new RBN is recorded in the RCT, whether or not the bad block has previously been replaced, as well as the bad block's old RBN if it was previously replaced, and the fact that the process is in phase two (which is indicated by flag P2) tstep 140); this information goes in sector zero of each RCT copy. The RCT is updated ~ithout reading sector zero, using instead the copy of RCT sector zero last read or written to the RCT. If the RCT cannot be updated, that i8 reported as an error to the error log (step 142) and con~rol of the replacement process jumps to step 166.
If there is no such jump, step 144 is performed next, and the process updates the RCT to indicate that ~5 the bad block has been replaced with the new RBN and that the old RBN, if any, is unusable. If this requires updating two blocks in the RCT, then both blocks are read before either is written. If a block cannot be read successfully, that error is reported to the error log (step 146) and control jumps to step 166. If a block cannot be rewritten successfully, that also is reported as an error to the error log (step 148) and the process jumps to step 162.
10In the absence of jumps or branches, step 150 is reached next. There, a replace command is used to alter ~- the header of the bad black, in indicate that it has been replaced in either the primary or secondary mode, and to contain 128 copies of the replacement block's RBN
address; and a write-compare command (addressed to the bad block's LBN) is used to store the saved data in the replacement block. If and only if the saved data is invalid, the write-compare is performed with the "forced error" modifier set. The replace command implicitly verifies that a header servo track failure has not occurred, which would cause a large number of improper replacements. If the replace command fails, Step 152, control branches to step 162. If the write command fails, Step 154, control branches back to step 136, to re-scan the RCT for another RBN. The current new RBN
will become the old RBN for this next pass. Either failure will already have been reported to the error log.
The write command succeeds if there i5 no error detected and the saved data is valid, or if only a forced error is detected and the saved data is invalid.
Next, in step 156, the process updates sector zero of the RCT copies to indicate that it is no longer in the middle of replacing a bad block. The RCT must be updated without reading sector zero, using instead the copy of sector zero last read ~rom or written to ~he RC~. If the RCT cannot be updated, the error is reported ~o the error log, step 158, and control branches to step 170.
In step 160, the process releases the lock it acquired in step 114 and ex;ts.
In step 162, the process res~ores the RCT to indicate that the new RBN is unallocated and usable and that the bad block is neither replaced nor has been revectored to the old RBN, whichever was it's original status. The RCT must be updated without reading any blocks from it, using instead the copies of the relevant blocks that were read from the RCT in step 144. Any errors are reported to the error log (step 164) but otherwise ignored.
Proceeding to step 166, the process uses a write command, addressed to the bad block's LBN, to restore the saved data. This write should also set the "forced error" flag if and only if the saved data is invalid.
Errors are reported to the error log, step 168, but otherwise are ignored.
The process next updates sector zero of the RCT
copies, step 170, to indicate that it is no longer in the middle of replacing the bad block. The RCT is updated without reading sector zero, instead using the copy of sector zero last read from or written to the RCT. Errors are reported to the error log, step 172, but otherwise are ignored.
In step 174, the process then releases the lock that it acquired in step 114. If a controller is performing the bad block replacement, it reports to the host the failed bad block replacement procedure; if a host is performing the bad block replacement, it takes whatever host-dependent action is appropriate for failed bad block 27 ~ 76~

replacement operation Step 176. That ends the process, and it exits.
When a disk is brought online to a host, the host or subsystem (whichever is performing bad block replacement), must do three things: (1) read sector zero of the RCT copies; (2) write the data just read back to sector zero of the RCT copies (this catches failures that occur in the middle of the multi-write routine); and (3) check the data just read to see if a failure occurred part way through bad block replacement (and, if so, resume the bad block replacement process as described above). Write access ~o the disk must not be allowed until after these actions have been performed and any partially completed bad block replacement has run to completion.
The foregoing algorithm guarantees tha~ da~a is never lost, as the best guess to the correct data is permanently stored before any action is taken that might destroy the data. There is a Pailure mode, however, in 2D host-initiated replacement, which assumes that the system crashes in the middle of bad block replacement. The bad block that has been partially replaced may be in the middle of the system core image or some other portion of the disk critical to booting the host system. This failure mode is eliminated if the subsystem, ra~her than the host, performs bad block replacement.
The Replacement and Caching ~ables The replacement and caching tables maintain a record of the locations of all revectored LBN sectors and the status of each RBN on the unit. Each RCT entry represents an RBN. In turn, each copy of the table has entries organized in ascending RBN order, with an entry for each RBN sector on the unit. There are "n" copies of the table on the unit, where "n" is a device characteristic. The tables are stored at the high address end of the LBN area of the unit. Table entries and RBN's are alloca~ed via a "hash" algorithm described elsewhere in this document.
A plurality of copies of the RCT are stored in the highest addresses of the LBN space. Each sector of the ~CT contains 128 entries, regardless of whether the disk is formatted as 512 or 576 bytes/sector. Each copy of the RCT is stored on an integral number of tracks. "Null entry" positions are added to adjust the RCT's size so that it meets this requirement. These null entries do not correspond to RBN's; there is always at least one null entry at the end of the RCT. Fig. 5 illustrates the format of a replacement block descriptor. There, 190 represents the lower order portion of a revectored LsN~s logical block number and 192 represents its higher order portion. A four-bit segment 194 is placed contiguous to that address. It is written with an octal status code.
Suitable exemplary codes and their meanings are listed below:
Code Meaning 00 unallocated (i.e., empty) replacement block 02 allocated replacement block-primary RBN
3003 allocated replacement block-non-primary RBN
04 unusable replacement block 3~
null entry Additionally, the size of the copies must be adjusted so that corresponding blocks of each copy are, to the maximum extent possible, accessed using -- ~9 1~'76~

physically distinct components. For conventionally structured devices, this implies. (1) if the number of copies is less than or equal to the number of read/write heads, then corresponding blocks of each copy are accessed by different heads; (2) if the number of copies is greater than the number of heads, then corresponding blocks of each copy are distributed as evenly as possible across all heads; (3) if a device uses a servo surface, then corresponding blocks of each copy are located using different tracks of the servo surface; and (4) the RCT
copies are allocated such that the last sector of the last copy occupies the last LBN on the unit. The last copy of the RCT is padded so that its size is an exact multiple of the device's track size. Allocation of the RCT is then performed starting at the highest LBN and working downward. The RCT pad area is controller-specific and is not accessed by the host.
The first sector in the RCT contains information about the state of any replacement operation which may be in progress. A copy of the volume serial number is also contained in this sectoF to allow validation of the RCT
by diagnostics routines.
The second sector in each copy of the RCT, sector 1, is used by the bad block replacement algorithm, as stated above. This sector is used to hold a copy of the data from the sector being replaced.
Fig. 6 illustrates the resulting RCT structure. The first sector 202 oE the RCT (i.e., the so-called sector 0) contains replacement and caching table information.
The second sector~ 204, (i.e., the so-called sector 1) contains the replaced LBN image. Sectors 206a-206m ~i.e., the so-called sectors 2 thru RCT-l) correspond to the 128 replacement block descriptors.

~ ~7~

And the ~ontents of sector zero of the RCT are illus-trated in Figures 7A and 7B. As sho~n there, the sector comprises 512 bytes. Words 260-266 contain the volume identification assig-ned during the formattin~ process. Word 268 contains four bits having individual significance. Bit 272 is the forced error (FE) Elag indicating that the replacement process should write the Eorced error indicator in the target RBN; as explained elsewhere herein, "setting" the forced error indication means writing the sector with the code for the forced error indicator, which in the example illustrated in the one's complement of the error detecting code for the sector. Bit 274 is a flag (BR0 for bad RBN's, in-dicating that the replacement in progr~ss was caused by a bad RBN;
it is cleared after the replacement process is over. Bits 275 and 276 are flags (P2 and Pl, respectively?, indicating whether a replacement is in progress and, if so, its phase. Words 278-282 hold a copy of the LBN of the block being replaced, if a replace-ment is in progress; this field is only valid when -the P2 bit 274 is set (i.e., when in phase 2 of replacement). Words 284, 286 contain a copy of the RBN of the block with which the LBN is being replaced~ if a replacement is in progress; they, too, re~uire that the RP flag be set. The RBN of the bad replacement block appears in words 288, 290 if the BR flag 274 is set.
The RCT replacement allocation algorithm is one which is used to allocate an RBN to replace a bad LBN. Described below are the search algorithm and various support algorithms~
_T Search The search begins at the descriptor for the primary i `; .r' replacement block. If the desired LsN address is not stored there and the descriptor is not empty, then a ping pong search begins of the sector containing the primary replacement block descriptor, If either the desired LsN address or an empty des-criptor is not encountered, then a linear scan of the remaining RCT blocks, and descriptors -30a-within blocks (with wrap-around at the end of the RCT), ensues until one of two things occurs: (1) an unallocated replacement block descriptor is encountered in an overflow location (a secondary) or (2) the entire RCT is searched without success (a failure).
The search operates at two levels. First, within the primary descriptor RCT sector, the search proceeds outward from the primary descriptor searched, starting with the next highest RBN descriptor. This degenerates to a linear search once the first or last descriptor is encountered. The linear search starts with the next highest RCT sector address, once the initial sector has been completely searched. Each new sector is searched in a linear fashion starting at the lowest RBN descriptor and scanning until the highest RBN descriptor in the sector is encountered. If at any time during the linear search a null (not an empty) entry is encountered, the search resumes at the first entry in the third RCT sector (the first with descriptors). The search is terminated when it is certain that all the RCT entries have been searched.
Fig. 8 illustrates the RCT search algorithm. A
listing of a sample coding for the algorithm is shown in Figs. 9A-9C.
The Primary RCT ~ash Alogrithm The primary RCT has algorithm is one which takes as inpu~ and LBN and produces a host LBN address of the RCT
block containing the primary R~N descriptor that corresponds to the previously revectored LBN. An offset pointing to the primary RBN descriptor within the RCT
block is also produced. This algorithm always produces a block number within the first copy of the RCT within the replacement control area. The algorithm is illustrated in Fig. 10.

6~

Detailed Description of an Embodiment of the Physîcal Layer With the foregoing generalized description in mind, it ~ill now be helpful to provide some :ELIrther details of an exemplary implementation.
Referring now to Figure 11, a suitable sector format is shown there, lllustrating the various sector fields: header 330, data 332, error detecting code tEDC) 334 and error correcting code (ECC) 336. Four copies of the logical address are provided within the header. The EDC in field 334 pro-vides error detection coverage from the entry of the data into the subsystem until its exit from the subsystem. It is also used in the illustrated embodi-ment to generate the "forced error indicator.7' Sixteen bits are used for the error detecting code in the present example, although codes of other lengths can be employed, of course. The ECC in field 336 provides the primary detection and correction mechanism against medium and device transmission errors. (An exemplary ECC occupies 170 bits and is described in commonly assigned United States Patent No. 4,413,339 titled Multiple Error Detecting and Correcting System Employing Reed-Solomon Codes).
The header preamble "spacer" field 338 is an area padded with zeroes and used to accommodate ~he maximum uncertainty between a drive's negation of sector pulse and a controller's notice of the change, plus the controller quantization error in preamble length.
The header preamble field 340, also zeroes, is the number of words necessary to allow the drive's phase-locking oscillator (PL0) to settle be-fore the ~.~.r - ` ` ~g~7~

occurrence of header syncO The "header preamble length" field is provided to the controller by the drive in response to a designa-ted command.
~eneration and Use of the Forced Error Indicator The controller 4 transmits to the drive 3 the informa-tion :Eor each sector, to be written in the format of Figure 11. Gen-erally, the error detecting code and error correcting code are computed and are inserted in the appropriate fields of the sector format by the controller 4, though in some cases the host or the drive itself might supply some of such information. Similarly, when reading from the disk drive 3, the controller 4 generally performs the functions of checking the error detecting code and error correcting code, though it is also possible :Eor the host or drive to do so in some systems.
The procedure by which the controller (or host or drive, as appropriate) generates and utilizes the error detecting code field 334 for forced error indication is illustrate~ as follows, for reading from the recording medium. Under command from the controller, the disk drive positions a head appropriately 50 as 2Q to "read" and provide to the controller the information stored in the sector to be read (hereafter "sector A"). From the data field of the sector, an error detecting code (EDC) is computed. The error detecting code calculated is then compared with the error detec-ting code contained in a predetermined field of the infor-mation read from sector A. If the calculated EDC matches the EDC
read from the recording medium, then the read operation is succes-sful and processing proceeds to an exit point. ~Iowever, if the ,~

two EDC values do not match, some further processing is required to determine the reason for the mismatch and to ~ecide on further steps to take. Thus, the controller next looks to see whether the forced error indicator was present (or "set") in sector A as recorded; this is done by comparing the calculated EDC with the one's complement of the EDC read from sector A, If the two match, that means the forced error indicator was detected, in which case processing continues; the controller then knows that the data in the sector already was corrupted when it was written and therefore cannot be recovered through the read operation.
Thus, the read operation fails; and in doing so, a specific code or signal is generated to notify the host that the read operation failed due to detection of the forced error indicator. On the other hand, if the forced error indicator was not detected, it is possible that error recovery technique might successfully be em-ployed, so control is transferred to a point where the available error recovery techniques are invoked~and an attempt is made to reconstruct the data written to sector A. If error recovery is successful, then the read operation has succeeded and control proceeds to an exit point, However, if error recovery does not succeed, then an error code is generated to indicate to the host that the read operation failed due to the medium inaccurately storing the information. This latter error code is different from the error code generated above, to distinguish between read errors due to defective media and read errors due to the forced error indicator.

-33a-.

$

It is important to distinguish between forced errors and non-forced errors in reading because they are handled dif-ferently when copying information from one sector or group of sec-tors on to another sector or group of sectors. When a forced error indicator has been set in a sector, the ~mderlying medium is still presumably in satisfactory condition; thus, new infor-mation can be written into that sector, at which point the forced error indicator is cleared, However, if the information in a sectox is unrecoverably corrupted but the forced error indicator is not present, it is likely that the medium underlying sector is defectlve and that any further writing onto that medium would result in a further loss of information.
There are thus three types of situations to deal with when copying a sector from a first volume of memory into a second volume of memory. If the sector was successfully read from the first volume of memory, it may of course be written intact into the second volume of memory. If the sector could not be succes-sfully read because it was logically corrupted and unrecoverable, then the contents of that sector can either be discarded entirely or can be written into a sector in the second volume with the for-ced error indicator set to signify that the data was corrupted when recorded. Rerecording of that sector subsequently results in the clearing of the forced error indicator. Thus the organ-ization of a particular file can be retained with the position in the file being held open and identified as not being defective though contalning incorrect data.

-33b-This is illustrated as follows, to which show the two possibilities for writing into the EDC/FEI fleld of sector A.
If a sector is to be written with information read from a dis~
file, and in the reading process the step has been reached which signifies that the data read is corrupted and the forced error indicator code is not present for the sector which was read, then the data known to be corrupted is to be written to a good block of recording medium on the same or another disk surface and the forced error indicator code is to be set when the new sector is written. This is indicated by branching directly to a step where the write operation occurs with the forced error indication code set in the appropriate field. By contrast, when the sector is to be written with new data from a host (which data is presumed to be reliable and uncorrupted3 or with data successfully read from storage, then the write operation occurs with an appropriate non-complemented error detection code written to the EDC/FEI field.
The data preamble "space" field 342 is the area neces~
sary to accommodate controller quantization errors in the tran--sition between reading headers and writing data preamble. The length of splice field 344 is the number of words necessary to accommodate worst-case header transmission delays, header compare time, write splice area and PLO lock time, The number for this area (in words) is placed in the "data preamble length" field of the response to the above-designated command.
The length of the write-to-read recovery field 346 is the number of bits necessary for write recovery, plus an allowance for uncertainty.

, . , -33c-The length of the reinst~uct time field 348 is the disk area traversed during the time the controller is cleaning up the current sector transfer and sending the command to the next one.
The Headers The sector header is 128 bits: thirty~two bits replica-ted follr times. The layout of one of the thirty-two bit copies is shown in Figure 12. A 16-bit word 352 and the lower 12 bits o~ the next word 354 form a 28-bit block number field, which is followed by a 4-bit header code 356. The block number field re-presents an LBN, an RBN, an XBN, or a DBN, depending on -the header code. The block number field provides enough addressing for approximately 0.25 giga-sectors or 1 terabit of data.
The octal header code may, for example, be interpreted as follows, First, an exemplary code such as 00 (octal) may in-dicate a usable logical sector wherein data may or may not be ~alid, depending upon the validity of the EDC. The block number in the header represents -33d-` ' ' ` ` .
.

3~ 7~

the LBN for this block. This header code may appear in LBN space only. Another code, such as 03, may indicate an unusable revectored logical sector. This header cod~
may appear in the non-RC~ portion of LBN space only. The data field con~ains the RBN header field of the replacement block, replicated 128 times; the block number in the header represents the LBN for this block.
Yet another code, such as 05, may indicate an unusable primary revectored logical sector. Such a sector has been revectored onto the first replacement sector on the track. The data field contains the RBN
header field of the replacement block, replicated 128 times. The block number in the header represents the LBN
for this block. This LBN has been revectored to its primary RBN. This header field may be registered in the non-RCT portion of LBN space only.
Yet another code, such as 06, may be used to indicate a usable replacement sector, wherein data may or may not be valid, depending on the validity of the EDC.
The block number in the header represents the RBN for this block. This header code may appear in RBN space only.
Another code, such as 11, may signify an unusable sector, where data is invalid. The block number in the header is that of the sector's type if it had been a usable sector. This header code may appear in RBN, XBN, or DBN space, the RCT area of LBN space, and in LBN's which have been secondary revectored due to header errors.
Yet another code, such as 12~ may signify a usable external sector, wherein data may or may not be valid depending upon the value of the EDC. The block number in the header represents the XBN for this block. This header code may appear in XBN space only. A further 35 ~

code, such as 14, may represent a usable diagnostic sector. The block number in the header represents the DBN for this block. This header code may appear in DBN
space only.
8eader Compare Al~orithm A header compare algorithm is used by the controller for locating a designated sector. First, the controller determines the address of the sector it is searching for on the disk ~i.eO, the "target" address). The controller then reads the four copies of the 32-bit header of the sector that may be at the target addressO These headers are broken into two 16 bit fields (low and high). If any two of the four low fields, as retrieved from the disk, match the low field of the target address and any two of the four high fields, as retrieved from the disk, match the high field of the target address, then the header compare succeeds. If at least two low matches are not found, then a header match is not possible.
If at least two low matches are found and two high matches are not found, then it iB possible that the correct sector was located but the header code did not match the target header code. ~his is possible if an LBN
has been replaced, or if a bad block has been found in a multi-copy protected area (i.e., RCT, XBN or D~N). The controller alters the header code in the target address then determines if two high matches now exist. A variant of the header compare algorithm is also used to conclude that a drive has mis-seeked or seeked to the wrong cylinder or group, or that an incorrect head has been selected. For this purpose, any three of the four high header words must match and any three of ~he four low header words must match, since there is not an expected header value to match against. Given this three~way match, the controller may interpret ~he header code and block number fields to de~ermine the actual cylinder, group and track that have been accessed, for comparison against the correct values.
The Data The contents of the data field are application-dependent. The data field size will depend on the format used by the host processor. For the assignee's products, there are two basic data field sizes, 512 bytes and 576 bytes. A por~ion of all disks is always formatted with 512 byte data fields. This is the manufacturing defect area (XBN). The other areas on disk drives attached to those controllers that support both sector sizes may be formatted in ei~her 512 or 576 byte format. Each time a device comes "online" to a controller, the controller is responsible for determining the sector size employed by the device according to the algorithm described below~
First, the device is instructed to change the sector size of its reading operation to 512 bytes. The star~ing sector of the first copy of format information is read.
The first word of this sector is tested. If it is equal to a preselected number, then LBN/RBN space is written in 512 byte mode. On the other handv if it is written with some other preselected number, then such space is written in 576 byte mode. If the starting XBN of the first copy is not readable or a value other than the aforementioned preselected values is in the first word, then the starting XBN of the next copy of the format control table is computed using the following Eormula:
next copy XBN = old copy XBN + size of format control table~
This new sector is ~hen read. If it has an uncorrectable I/O error, then the next copy is accessed, until all copies are tried. If all copies are read and there is no copy that can be read without an uncorrectable I/O error, 37 ~ S

then a media format error is returned to the host. Also, if the first sector (i.e., XBN) of the first copy read without an uncorrec~able error contains an invalid media mode code, then a media format error is returned to the host.
(The host may force the device into a speciEic mode, in which case the controller will attempt to access the device unit using that mode, without issuing the media format error. This is intended only as a means of data recovery, and not as a standard operating practice.) If the volume is in 512 byte formatt the algorithm is complete. If in 576 by~e forma~, the controller is responsible for prefacing all operations on X~N's or 512 byte DBN's with a command to change the size to 512 bytes, and preceeding the next reference to LBN'S or RBN's with a command to change the size back to 576 byte format. In other words, the controller is responsible for changing the sector size dynamically based on which space the sector falls in, using 512 byte format for ~0 XBN's and DBN's but 576 byte format for LBN's and RBN's.
The EDC
The Error Detecting Code (EDC~ is a 16 bit code used to detect errors caused by internal problems in the controller. It is applied as an end-to-end verifica~ion of correct controller operation. The algorithm shown here was designed to detect column errors as well as multi-bit parity errors.
The EDC is computed via an exclusive-OR operation and left circular shift algorithm, using a non-2ero initial value and 16 bit word size. The rotate used in this algorithm has no carry. The algorithm itself is listed in Fig. 13. In addition to detecting errors, the EDC also is used herein to provide a forced error indicator. This is accomplished by storing the one's 38 ~8~7Ei~5;

complement of the correct EDC in the EDC field of the sector. An 'lerror" is thereby indicated when the sector is read; this "error" is eliminated when the sector is next written with correct EDC. This technique makes it very easy for diagnostic routines to identify sectors having forced errors. That is, when an EDC indicates an error, it is a simple matter to determine whe~her that EDC is in fact the one's complement o~ the EDC to be expected on the basis of the recorded data.
The Track A track is composed of sectors and timing marks.
There must be at least two sectors per track (1 LBN
sector and 1 ~BN sector), Timing marks are of two types:
(1) sector marks and (2~ index marks. A sector mark precedes each sector and may be used by the controller for rotational optimization purposes. An index mark precedes the first sector on each track within the first group in the cylinder and precedes a sector at the same angular position wi~h respect to the first group on al]
~0 other tracks within all other groups in a cylinder.
Detailed Description of an Embodiment of The Logical Layer Address SPaceS
There are four address spaces in the set of sectors made available to the controller by the drive. The first address space contains the set of logical blocks which are visible to the host~ ThiS LBN space is divided into two regions: the host acce~sible area and the RCT's.
The second address space contains replacement blocks which are used to replace logical blocks ~ha~ have become unusableO These RB~'s are invisible to the host except for the implications they have on allocation policies.
The controller utilizes the logical blocks and 3g ~

replacement blocks in a fashion that presents to the host a logically contiguous se~ of blocks numbered from zero to H-l, where H is the block capacity as seen from the host. The third address space i5 the extended block space; (XBN's); this is a set of blocks visible only to the controller, which is used to store manufacturing ormat control information and transient controller-specific information. Finally, there is the diagnostic block space (DBN's) containing blocks devoted to controller-resident diagnostics. The DBN'S are also visible only to the controller~ These address spaces are differentiated by unique header codes, preventing inadvertent access to or operation in the wrong type of sector.
Although conformation to the overall geometry described herein is a requirement of the invention, the specific capacities and other physical parameters associated with the geometry of the disk will vary from device type to device type. These specific parameters are part of the permanent characteristics of each device type, and are determined when the device is designed~
The controller shields from the host these parameter-dependent device properties. The controller issues a generic command termed the GET CHARACTERISTICS
command, in response to which the drive responds by sending to the controller the parameters necessary for use in geometry-related operations. The con~roller then uses those parameters as appropriate and necessary.
The Drive Characteristics Blocks _ As mentioned above, in a secondary storage subsystem according to this invention, a disk drive provides to the controller, responsive to a command, one or more messages containing various parametric informationO In this regard, it should be noted that within a drive ~here may ~o~7~

be one or more subunits~ each of which can be addressed independently by the host and control~er. ThuS, to fully characterize the drive, two commands are used. First, a command named the GET COMMON CHARACTERIS~ICS command is employed to evoke a message regarding parameters which are common to all subunits of the drive. ~ext, a comand named the GET SUBUNIT CHARACTERISTICS command is used to evoke the characteris~ics of specific subuni~s of a drive. The format of the response to the GET COMMON
CHARACTERISTICS command is illustrated in Fig. 14.
There, a 23 byte sequence is shown. The first byte identifies the nature of the response. The lower half of the second byte conveys the length of a short time out, expressed as power of two. The upper half of the second byte contains a number indicating in the version of the bus used between the controller and drive. In the third byte, the drives bit transfer rate is specified, scaled down by a factor of 100,000. The fourth bytel like the second byte, is broken in half. Its lower half includes a long time out, also expressed as a power of two; while its upper half conveys the number of retries of a failed operation which will be reguired by the drive. In the lower half of the fifth byte, a number is written to indicate the number of FCT and RCT copies ~aintained.
m e most significant bit in the fifth byte, SS, indicates the drive sector size. The sixth byte specifies the number of error recovery levels which the drive makes available. It is a characteristic of this system that the controller need not be aware of the error recovery techniques available in the drive. The drive may employ several different error recovery techniques, numbered in their order of increasing or decreasing chance of success. Assume, for example, that by convention error recovery Level 1 corresponds to the technique having the 4~ 76~;

greatest probability of success; error recovery Level 2 is the neYt most likely to succeed, etc. Then, the controller need only signal for the invocation of error recovery Level 1 and the subsequent error recovery techniques, in ascending numerical order (corresponding to descending probability of success). The drive, responsive to seeing each o~ the error recovery level indicators, invokes the appropriate r~overy method.
~he seventh byte contains the BCC threshold, above which replacement and revectoring are invoked~ The eighth byte contains an indication of the microcode revision number of the drive and the ninth byte contains an indication of its hardware revision number~
Bytes lO 15 contain a unique drive identification number or serial number. The sixteenth by~e contains a drive type identifier and byte seventeen indicates the rotational speed of the disk platters, in revolutions per second.
Bytes 18-23 contain various error thresholds.
The response to the GET SUBVNIT CHARACTERISTICS
command is indicated in Figs. 15A and 15B. As shown there, the response is 39 bytes in length. The first byte contains a pattern indicating the nature of ~he response. aytes 2, 3, 4 and the lower-order half of byte 5 contain the number of cylinders in the hBN space. The field comprising bits 6-4 of byte 5 contains bits number 30-28 of all cylinder numbers on ~his subunit.
The number of groups per cylinder is indicated in byte 6.
The lower-order half of byte 7 contains bits 27-24 of the first LBN on this subunit, while the upper-order half of that byte contains the same bits of the first XBN
on this subunit. Byte 8 contains ~he number of tracks per group. Byte 9 is fragmented into two halves, the ` 42 ~76~

lower half of which contains the bits 27-24 of the first RBN on this subunit, while the upper half of the byte contains the same bits for the first DBN on this subunit.
The number of RBN's per track is indicated on byte 10. Bytes 12 and 13 con~ain the length of the data and header preambles, respectively, in words.
Bytes 14-17 record the media type. Bytes 18 and 19 give the size of copy of the FCT, in XBN'S.
Bytes 20-27 are used for the 512 by~e format, and their counterpart for the 576 byte format is bytes 28-35.
As labelled in the drawing~ the contents of the bytes should be self-explanatory. Bytes 20 and 28 indicate the number of LB~'s per track. Bytes 21 and 29 indicate the group offset - i.e., the offset from one group to another to permit spiral read operation. The number o LBN's in the host area is indicated from byte 22 through the lower-half of byte 25 and from byte 30 to the lower-half of byte 33. Bytes 20-23 and 34-35 indicate the size of a copy of the RCT~ in LBN's.
Bytes 36-39 are common to both formats. Bytes 36 and 37 indicate the size of the XBN space, in cylinders.
Byte 38 indicates the number of groups in the DBN area and byte 3g indicates the size of the DBN space in cylinders.
The replacement sectors in any given drive are logically numbered from 0 ~o (Rs - 1), where Rs =
Lc*g*t*r is the total number of replacement sectors. A
replacement block number is conveeted to a speciic physical disk location ~hrough a series of transformations performed by the controller using parameters supplied by the drive. These transformations are described later. The last r sectors (where r is a drive~specific parameter) of each track in the host application area is reserved for replacement blocks for 43 ~7~

revectored bad blocks. These alterna~e blocks lie outside of the LBN space pre~ented by the controller to the host, and are accommodated in ~he logical to-physical address conversion algorithm described below.
Fig. 16 illustrates the first two and last tracks in the LBN/RBN space of a subunit.
External Block Track Geometry The external sectors on any given drive are logically numbered from (0 to Xtot - 1), where Xtot =
Xc*g*t*s and is the total number of external sectors.
The transformation for ccnverting an external block number ~o a specific physical disk location is explained later.
XBN's are allocated contiguously on all XBN
cylinders; they increase incrementally from the starting XBN number as the sector number r track number, and cylinder number increase, un~ he XBN cylinders are exhausted. There are no replacement blocks on XBN
cylinders.
Fig. 17 illustrates the first two and last track in the XBN space of a subunit.
Diagnostic C~linder Geometry The diagnostic sectors on a drive are numbered logically from 0 to Ds-l, where Ds - Dc*g*t*s is the total number of diagnostic sectors. The method for transforming DBN's to specific physical disk locations is described below. An adequate number of cylinders is reserved for diagnostic usage~ Sector headers in those cylinders are coded to reflect that they are DBN's~
These diagnostic cylinders are formatted initially in the 512 byte mode and the last cylinder in this space must remain in that mode; that cylinder contains various data patterns prerecorded at the factory. Diagnostic space geometry is illu~trated in Fig. 18.

~ 44 ~76~S

Address Conversions ~ wo generic variables are used to express the address conversion algorithms. They are actual or calculated device characteristics, The functio~ QUO( ) is used to indicate a quotient resulting from a division operation and the function REM( ) is used to indicate the remainder resulting from a division operation.
The starting LBN for a drive (L) is computed from the characteristic "HISTRTLBN"~ the high order part of the address of the starting LBN~ (see below). This is done by OR-ing the "HISTRTLBN" nibble into bits 27-24 of a previously zeroed longword.
Given a header LBN; the algorithm listed in Fig. 19 is used to determine the logical block's physical sector address. In reading that figure, note that the starting cylinder for a drive (C) is computed from the drive characteristic "HI CYL", the high order part of ~he cylinder address. This i5 done by OR-ing the "HI C~L'I
nibble into bits 30-2~ of a previously zeroed longword.
In the figure, llo'l represents an offset.
Given a header RBN, the algorithm of Fig. 20 may be used to determine the replacement block's physical sector address. Note that the starting RBN for a drive (R) is computed from the characteristic "HISTRTRBN," the high order part of the RBN address. This is done by O~~ing ~he "~ISTRTRBN" nibble into bits 27-24 of a previously zeroed longword.
Given a header XBN, the algorithm listed in Fig. 21 may be used to determine the external block's physical sector address. The starting XBN for a drive (X) is computed from the drive characteris~ic "~ISTRTXBN," the high order part of the XBN addre~s. rhis is done by OR-ing the ~HISTRTXB~" nibble into bits 27-24 of a previously zeroed longword.

~s Given a header DBN, the controller executes the algorithm of Fig. 22 to determine the diagnostic block's physical sector address. The starting DBN for a drive (D) is computed from the characteristic "HISTRTDB~," the high order part of the DBN address. ~his is done by OR-ing the "HISTRTDBN" nibble into bits 27-24 of a previously zeroed longword.
Given a header LBN that has been revectored to the first R~N on the same track (primary RBN), then the following algorithm or formula may be used to determine the replacement block's RBN:
RBN = R + (QUO((LBN - L)/l))*r Given a host LBN that has been revectored to the first RBN on the same track (primary RBN), then the following formula may be used to determine the replacement block's RBN:
RBN = R + (QUO((LBN)/1)3*r Given the physical address (cylinder~ group and track) of a logical block that has revectored to the first RBN on the same track (primary RBN), then the following formula may be used to determine the replacement block's RBN:
RBN = R + ([([(Cyl. No. -C)*g] + Group No.)*t] + Track No,)*r Detailed Description of an Embodiment of the Functional L~
Revector'ng Once a sector has been replaced, revectoring should occur upon each access to the replaced LB~, Three revectoring mechanisms are supported by the particular implemention discussed herein. These mechanisms all depend upon values in the code field of the sector's header to initiate revectoring. ~dditionally~ all revectored LBN's contain 128 copies oE the replacement ~376~a$

block's header in their data field, unless revectoring is the result of a header error. The revectoring mechanism6 differ in the ways that the addresses of the target RBNIs are determined.
In the primary revectoring mechanism, the position oE the RBN to which revectoring is performed is implied by the position of the LBN on the volume. This implied position is the first replacement sec~or on the track containing the LBN. This is a many ~BN ~o one ~BN
mapping function.
With so-called secondary revec~oring, an arbitrary RB~ is used whose address is determined by the presence of the 128 copies of the RBN's header value (code and address) in the data field of the bad LBN. The algorithm listed in Fig. 23 is used to determine reliably the correct value of the RBN header; it provides as output (from the 128 copies input~ the address ~ound to have at least 24 matches, if there is one.
Finally, there is a so-called tertiary revectoring mechanism which is used when the header compare algorithm fails to determine a valid header address or code or the algorithm of Fig. 23 fails to yield a valid result. It is important to determine then if the LBN has been revectored ox if access to the LBN should result in an unrecoverable error. Since all revectored LBN's are recorded in the multiple copies of the RCT, an RCT search is used to determine if the bad LBN has been revectored.
The RCT search algorithm, described above, results in the RBN address if the LBN was revectored, or a failure indication if it was not revectored. The determination that the attempted input/output operation was done to the correct sector requiresl since the header is n smashed"
and unusable ~1) a determination that ~he correct cylinder, group and ~rack have been selec~ed; (2~ for 47 ~ 6~5 controllers tha~ use sector counting via ~ector and index pulses, at least one revolution of counting after completion of the foregoing step and (3) for controllers that locate sectors by reading headers, a~ least four full revolutions searched after the ~oregoing step is complete. Failure to achieve a header match on the latter two actions requires invocation of tertiary revectoring.
F _ atting Support Formatting and reformattin~ processes are responsible for establishing which sectors are bad and replacing them, if they are in the host applications area, or formatting there headers with the unusable code if they are bad LB~'s in the RCTy bad XBN's, bad DBN's or bad RBN's.
The formatting process is supported by the format control tables ~FC~), which are used to record information about the location of manufacturing detected bad blocks. Format information for both 512 byte and S76 byte formats is stored in the FCT. The first subtable in the FCT contains information about where the bad blocks would be located if the disk were located in the 512 byte ormat, the second subtable contains information about where the bad blocks would be located if the disk were recorded in the 576 byte format. For those mass s~orage devices that don't support the 576 byte format, the 576 byte subtable contains null entries.
A second function of the FCT is the identification of the current mode of the LBN space (i.e., whether it is recorded in 512 or 576 byte format. The first sector of each FCT copy contains a code identifying the current LBN
sector size. This mode identification sector is updated each time the volume i5 formatted.

The FCT contains at least one track of subsystem scratch storage also.
Each copy of the FCT is composed of one volume information block, one 512 byte ~ormat table, one 576 byte format table, and one subsys~em temporary storage area (distrib~ted amongst ~he alignment pads). ThiS
format is illustrated in ~ig. 24. The XBN area itself is always for~atted to contain 512 byte sectors. Sector 0 of the FCT contains various volume identification 0 information. Its ormat is illustrated in Fig. 25.
C nclusion Having thus described an exemplary embodiment of the invention, it will be apparent that various alterations, modifications and improvements will readily occur to those skilled in the art. Such obvious alterations, modifications and improvements, though not expressly described above, are nonetheless intended to be implied and are within the spirit and scope of the invention~
Accordingly, the foregoing discussion is intended to be illustrative only, and not limitin~; the invention is limited and defined onIy by the following claims and equivalents thereto.

Claims (16)

THE EMBODIMENTS OF THE INVENTION IN WHICH AN EXCLUSIVE
PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:
1. In a secondary storage subsystem for a data processing system, wherein data is recorded on a mass storage medium and the smallest addressable unit of the medium is a sector, each sector including a header field for recording address information and a data field for recording data to be associated with and stored at such address, the improvement comprising:
means for writing in each sector a predetermined code, termed the forced error indicator, when the data being recorded in the data field is known to be logically corrupted and the medium underlying the sector is not known to be defective; and means for providing a signal, termed a forced error signal responsive to detection of the forced error indicator during a read operation.
2. The apparatus of claim 1 wherein the means for writing the forced error indicator code includes means for generating for each sector an error detection code which is uniquely related to such sector's data, in accordance with a preselected algorithm, for use in detecting the presence of errors in reading data recorded in the sector; and means for writing a signal into a predetermined location in the sector, such location being termed the EDC/FEI field, said signal being (a) the error detection code for the data being recorded in the sector when the data is not known to be logically corrupted, or (b) the forced error indicator code when the data being recorded in the sector is known to be logically corrupted.
3. The apparatus of claim 2 wherein the forced error indicator code is the one's complement of the error detection code for the data being recorded in the sector.
4. The apparatus of claim 3 further including:
means for generating a second error detection code upon reading the data recorded in a sector, using the preselected algorithm; and the means for providing a forced error signal comprising means for comparing the error detection code thus generated with the signal read from the EDC/FEI field and for generating the forced error indicator when the signal read from the EDC/FEI field corresponds to the one's complement of the second error detection code.
5. In a secondary storage subsystem for a data processing system, wherein data is recorded on a mass storage medium and the smallest addressable unit of the medium is a sector, a method of replacing a defective sector with a substitute sector, such that information to be written to or read from a defective sector is written to and then read from the substitute sector instead, once the defective sector is identified as unreliable, such method comprising the steps of:
A. reserving a portion of the medium to be used as spare sectors for replacing defective sectors, at least one spare sector being provided within each set of sectors occupying contiguious logical locations on the medium;

B. replacing the first defective sector in said set of sectors with the first one of said spare sectors, termed a primary replacement sector;

C. indicating such replacement by writing a first predetermined code in the defective sector; and D. when writing data to or reading data from a sector, detecting the first predetermined code and, in response to detecting said code, revectoring the writing or reading operation to said first one of the spare sectors for the involved set of sectors where the code was detected.
6. The method of claim 5 wherein each sector comprises a header field and a data field, the address of the sector normally being written in the header field and the information to be stored in the sector being written in the data field, and further wherein said predetermined code is written in the header field of the defective sector.
7. The method of claim 6 wherein the reserved sectors are evenly distributed throughout the medium.
8. The method of claim 6 wherein the medium is a magnetic disk and the sets of sectors are tracks.
9. The method of claim 6 wherein the reserved sectors are located in predefined locations within said tracks.
10. The method of claim 6 further including the steps of, when the primary replacement sector is unavailable:

E. selecting for a defective sector other than the first defective sector in said set of sectors a replacement sector other than the primary replacement sector, said replacement sector being termed a secondary replacement sector;
F. in the header field of such defective sector writing a second predetermined code indicating that said sector has been replaced by a secondary replacement sector;
G. in the data field of each such defective sector, writing a pre-determined multiple number of copies of the physical address of the secondary replacement sector selected therefor;
H. on reading the header field of the defective sector, checking for said second code;
I. responsive to detecting said second code, obtaining the physical address of the secondary replacement sector by reading said multiple copies and comparing them statistically to arrive at the recorded value of said address;and J. revectoring the writing or reading operation intended for the defective sector to said secondary replacement sector.
11. The method of claim 10 wherein the reserved sectors are evenly distributed throughout the medium.
12. The method of claim 10 wherein the medium is a magnetic disk and the sets of sectors are tracks.
13. The method of claim 10 wherein the primary replacement sectors are located in predefined locations within said tracks.
14. The method of claim 10 further including the steps of, when the prim-ary replacement sector is unavailable:
K. providing on the medium multiple copies of a table containing a list of each spare sector and the address of the defective sector replaced by it, if any;
L. responsive to detecting a defective sector, searching a copy of said table to find the address of the replacement sector therefore; and M. revectoring the writing or reading operation intended for the defective sector to said replacement sector.
15. In a secondary storage subsystem for a data processing system,wherein data is recorded on a mass storage medium and the smallest addressable unit of the medium is a sector, each sector comprising a header field and a data field, the address of the sector normally being written in the header field and the information to be stored in the sector being written in the data field, a method of replacing a defective sector with a substitute sector, such that information to be written to or read from a defective sector is written to and then read from the substitute sector instead, once the defective sector is identified as unreliable, such method comprising the steps of:
A. reserving a portion of the medium to be used as spare sectors for replacing defective sectors;
B. selecting for a defective sector a replacement sector from among said spare sectors;
C. in the header field of such defective sector, writing a predeter-mined code indicating that said sector has been replaced by a secondary replace-ment sector;

D. in the data field of a defective sector, writing a predetermined multiple number of copies of the physical address of the selected replacement sector therefor, termed a secondary replacement sector;
E. on reading the header field of the defective sector, checking for said second code, F. responsive to detecting said second code, obtaining the physical address of the secondary replacement sector by reading said multiple copies and comparing them statistically to arrive at the recorded value of said address;and G. revectoring the writing or reading operation intended for the defective sector to said secondary replacement sector.
16. In a secondary storage subsystem for a data processing system, wherein data is recorded on a mass storage medium and the smallest addressable unit of the medium is a sector, a method of replacing a defective sector with a substitute sector, such that information to be written to or read from a defective sector is written to and then read from the substitute sector instead, once the defective sector is identified as unreliable, such method comprising the steps of:
A. reserving a portion of the medium to be used as spare sectors for replacing defective sectors;
B. selecting for a defective sector a replacement sector from among said spare sectors;
C. providing on the medium multiple copies of a table containing a list of each spare sector and the address of the defective sector replaced by it, if any;
D. responsive to detecting a defective sector, searching a copy of said table to find the address of the replacement sector therefor; and E. revectoring the writing or reading operation intended for the defective sector to said replacement sector.
CA000412883A 1981-10-05 1982-10-05 Disk format for secondary storage system Expired CA1187615A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CA000475224A CA1203618A (en) 1981-10-05 1985-02-26 Disk format for secondary storage system
CA000475223A CA1205193A (en) 1981-10-05 1985-02-26 Disk format for secondary storage system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US06/308,771 US4434487A (en) 1981-10-05 1981-10-05 Disk format for secondary storage system
US308,771 1989-02-09

Related Child Applications (1)

Application Number Title Priority Date Filing Date
CA000475224A Division CA1203618A (en) 1981-10-05 1985-02-26 Disk format for secondary storage system

Publications (1)

Publication Number Publication Date
CA1187615A true CA1187615A (en) 1985-05-21

Family

ID=23195330

Family Applications (1)

Application Number Title Priority Date Filing Date
CA000412883A Expired CA1187615A (en) 1981-10-05 1982-10-05 Disk format for secondary storage system

Country Status (6)

Country Link
US (1) US4434487A (en)
EP (3) EP0090040B1 (en)
JP (6) JPH0810535B2 (en)
CA (1) CA1187615A (en)
DE (1) DE3280359D1 (en)
WO (1) WO1983001334A1 (en)

Families Citing this family (96)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4598357A (en) * 1980-11-14 1986-07-01 Sperry Corporation Cache/disk subsystem with file number for recovery of cached data
JPS5860410A (en) * 1981-10-06 1983-04-09 Mitsubishi Electric Corp Magnetic disk control system
CA1203019A (en) * 1982-01-19 1986-04-08 Tetsu Watanabe Apparatus for recording and reproducing a digital signal
JPS58194143A (en) * 1982-05-07 1983-11-12 Hitachi Ltd Recording and reproducing system of data
JPS59165162A (en) * 1983-03-11 1984-09-18 インタ−ナシヨナル ビジネス マシ−ンズ コ−ポレ−シヨン Volume restoration system
JPS59214952A (en) * 1983-05-20 1984-12-04 Nec Corp Processing system of fault
DE3481659D1 (en) * 1983-05-23 1990-04-19 Data General Corp METHOD AND CONTROL DEVICE FOR TREATING ERRORS OF THE RECORDING CARRIER IN A DISK STORAGE SYSTEM.
JPH0634304B2 (en) * 1983-09-01 1994-05-02 ソニー株式会社 Disk playback device
NL8303765A (en) * 1983-11-02 1985-06-03 Philips Nv DATA PROCESSING SYSTEM IN WHICH MEMORY UNRELIABLE WORDS ARE REPLACED BY AN UNRELIABILITY INDICATOR.
FR2561428B1 (en) * 1984-03-16 1986-09-12 Bull Sa DISC MEMORY RECORDING METHOD AND DISC MEMORY SYSTEM
DE3575646D1 (en) * 1984-03-24 1990-03-01 Philips Nv METHOD FOR TRANSMITTING INFORMATION WITH ERROR CORRECTION FOR DATA WORDS, AN ERROR CORRECTION DECODING METHOD FOR SUCH DATA WORDS, AN ARRANGEMENT FOR INFORMATION TRANSFER FOR USE WITH THE METHOD, AND A METHOD FOR USING AN APPARATUS.
US4631723A (en) * 1984-06-08 1986-12-23 Honeywell Information Systems Inc. Mass storage disk drive defective media handling
US4928193A (en) * 1984-07-13 1990-05-22 International Business Machines Corporation Diskette drive type determination
US5202979A (en) * 1985-05-08 1993-04-13 Thinking Machines Corporation Storage system using multiple independently mechanically-driven storage units
JPH0756734B2 (en) * 1985-05-27 1995-06-14 松下電器産業株式会社 Information recording / reproducing device
EP0238318B1 (en) * 1986-03-19 1992-03-04 Fujitsu Limited Track access control system for magnetic disk system
JPS6364678A (en) * 1986-09-05 1988-03-23 Mitsubishi Electric Corp Write error recovery system for disk
US4953122A (en) * 1986-10-31 1990-08-28 Laserdrive Ltd. Pseudo-erasable and rewritable write-once optical disk memory system
US4821253A (en) * 1986-12-19 1989-04-11 Matsushita Electric Industrial Co., Ltd. Optical disk and optical disk apparatus with error correction
DE3787452T2 (en) * 1986-12-19 1994-01-13 Matsushita Electric Ind Co Ltd Erasable optical disc and device for recording and reproducing optical information with device for handling defective sectors.
US4833663A (en) * 1987-01-16 1989-05-23 Matsushita Electric Industrial Co., Ltd. Information recording/reproducing apparatus for handling defective sectors on an optical disk
DE3884018T2 (en) * 1987-03-10 1994-01-27 Toshiba Kawasaki Kk Device for recording data on an optical recording medium.
US4858034A (en) * 1987-05-20 1989-08-15 Plus Development Corporation Modular unitary disk file subsystem with differing density zones
WO1989000312A1 (en) * 1987-07-02 1989-01-12 Exabyte Corporation Method and apparatus for data buffer management
US4811124A (en) * 1987-07-24 1989-03-07 Advanced Micro Devices, Inc. Defect skipping mechanism for disk drives
US4873688A (en) * 1987-10-05 1989-10-10 Idaho Research Foundation High-speed real-time Reed-Solomon decoder
JPH01130243A (en) * 1987-11-17 1989-05-23 Hitachi Ltd Fault recovering system for storage device
US5129088A (en) * 1987-11-30 1992-07-07 International Business Machines Corporation Data processing method to create virtual disks from non-contiguous groups of logically contiguous addressable blocks of direct access storage device
EP0401258B1 (en) * 1988-02-11 1994-11-30 VOGEL, Peter Samuel Document marking system
US5146571A (en) * 1988-03-28 1992-09-08 Emc Corporation Remapping defects in a storage system through the use of a tree structure
EP0365634B1 (en) * 1988-04-08 1993-10-20 Digital Equipment Corporation Method and apparatus for encoding consisting of forming a codeword by combining a first code sequence with a second code sequence
US5237574A (en) * 1988-04-08 1993-08-17 Digital Equipment Corporation Error-resilient information encoding
US5113512A (en) * 1988-06-21 1992-05-12 Matsushita Electric Industrial Co., Ltd. System for managing a storage medium reducing physical space needed
US4974189A (en) * 1988-08-16 1990-11-27 Hewlett Packard Company Magnetic tape packet assembler/disassembler safeguards existing data with pretries during appends
EP1031992B1 (en) 1989-04-13 2006-06-21 SanDisk Corporation Flash EEPROM system
US5083264A (en) * 1989-04-24 1992-01-21 Xerox Corporation Process and apparatus for saving and restoring critical files on the disk memory of an electrostatographic reproduction machine
US5200864A (en) * 1989-06-28 1993-04-06 International Business Machines Corporation Combining small records into a single record block for recording on a record media
GB8915875D0 (en) * 1989-07-11 1989-08-31 Intelligence Quotient United K A method of operating a data processing system
US5072378A (en) * 1989-12-18 1991-12-10 Storage Technology Corporation Direct access storage device with independently stored parity
US5737632A (en) 1989-12-19 1998-04-07 Hitachi, Ltd. Magnetic disc control apparatus with parallel data transfer between disc control unit and encoder/decoder circuit
US5367652A (en) * 1990-02-02 1994-11-22 Golden Jeffrey A Disc drive translation and defect management apparatus and method
US5233618A (en) * 1990-03-02 1993-08-03 Micro Technology, Inc. Data correcting applicable to redundant arrays of independent disks
US5195100A (en) * 1990-03-02 1993-03-16 Micro Technology, Inc. Non-volatile memory storage of write operation identifier in data sotrage device
US5083229A (en) * 1990-04-16 1992-01-21 International Business Machines Corporation Disk drive with minimized head-arm excursions
US5271018A (en) * 1990-04-27 1993-12-14 Next, Inc. Method and apparatus for media defect management and media addressing
JP2836929B2 (en) * 1990-07-05 1998-12-14 株式会社日立製作所 Rotary storage device and control method thereof
US5066749A (en) * 1990-09-11 1991-11-19 National Starch And Chemical Investment Holding Corporation Hydrophobically-modified polycarboxylates and process for their preparation
US5166935A (en) * 1990-09-28 1992-11-24 International Business Machines Corporation Method of determining correctness and contents of control data structures in moving media data storage systems
JP2761289B2 (en) * 1990-11-30 1998-06-04 富士通株式会社 Disk track emulation method
WO1992017836A1 (en) * 1991-03-29 1992-10-15 Sharples Kenneth R Electronic floppy disk emulation system
US5249288A (en) * 1991-04-01 1993-09-28 Xerox Corporation Process for accommodating bad disk pages in an electronic printing system
US5216655A (en) * 1991-06-26 1993-06-01 Digital Equipment Corporation Method and apparatus for surface reallocation for improved manufacturing process margin
US5303219A (en) * 1991-09-09 1994-04-12 International Business Machines Corporation Reclamation of dust contaminated sectors in optical disk apparatus
JPH05181617A (en) * 1991-12-26 1993-07-23 Hitachi Ltd Improvement of reliability for disk subsystem
US5519849A (en) * 1992-12-07 1996-05-21 Digital Equipment Corporation Method of reducing the complexity of an I/O request to a RAID-4 or RAID-5 array
US5819109A (en) * 1992-12-07 1998-10-06 Digital Equipment Corporation System for storing pending parity update log entries, calculating new parity, updating the parity block, and removing each entry from the log when update is complete
US6269453B1 (en) 1993-06-29 2001-07-31 Compaq Computer Corporation Method for reorganizing the data on a RAID-4 or RAID-5 array in the absence of one disk
US5504858A (en) * 1993-06-29 1996-04-02 Digital Equipment Corporation Method and apparatus for preserving data integrity in a multiple disk raid organized storage system
US5581690A (en) * 1993-06-29 1996-12-03 Digital Equipment Corporation Method and apparatus for preventing the use of corrupt data in a multiple disk raid organized storage system
US5522031A (en) * 1993-06-29 1996-05-28 Digital Equipment Corporation Method and apparatus for the on-line restoration of a disk in a RAID-4 or RAID-5 array with concurrent access by applications
US5390327A (en) * 1993-06-29 1995-02-14 Digital Equipment Corporation Method for on-line reorganization of the data on a RAID-4 or RAID-5 array in the absence of one disk and the on-line restoration of a replacement disk
US5564011A (en) * 1993-10-05 1996-10-08 International Business Machines Corporation System and method for maintaining file data access in case of dynamic critical sector failure
US5442638A (en) * 1994-01-10 1995-08-15 International Business Machines Corporation Apparatus and method for recording over defects in storage media
US5548795A (en) * 1994-03-28 1996-08-20 Quantum Corporation Method for determining command execution dependencies within command queue reordering process
US5696775A (en) * 1994-09-23 1997-12-09 Cirrus Logic, Inc. Method and apparatus for detecting the transfer of a wrong sector
US5649152A (en) * 1994-10-13 1997-07-15 Vinca Corporation Method and system for providing a static snapshot of data stored on a mass storage system
US5835953A (en) * 1994-10-13 1998-11-10 Vinca Corporation Backup system that takes a snapshot of the locations in a mass storage device that has been identified for updating prior to updating
US5826001A (en) * 1995-10-13 1998-10-20 Digital Equipment Corporation Reconstructing data blocks in a raid array data storage system having storage device metadata and raid set metadata
US5774643A (en) * 1995-10-13 1998-06-30 Digital Equipment Corporation Enhanced raid write hole protection and recovery
US5933592A (en) * 1995-10-13 1999-08-03 Digital Equipment Corporation Promoting device level error to raidset level error to restore redundacy in a raid array data storage system
US6334195B1 (en) * 1995-12-29 2001-12-25 Lsi Logic Corporation Use of hot spare drives to boost performance during nominal raid operation
US5721816A (en) * 1996-07-29 1998-02-24 Kusbel; Paul F. Adaptive recovery of read and write errors in a disc drive
JP3674227B2 (en) * 1997-03-14 2005-07-20 株式会社日立製作所 Storage device for storing portable media
JPH10320913A (en) * 1997-05-23 1998-12-04 Sony Corp Apparatuses and method for recording data, for reproducing data, for recording reproducing data, and transmission medium
US6038676A (en) * 1997-09-25 2000-03-14 International Business Machines Corporation Method and circuit for data integrity verification during DASD data transfer
US6370605B1 (en) * 1999-03-04 2002-04-09 Sun Microsystems, Inc. Switch based scalable performance storage architecture
US6247152B1 (en) 1999-03-31 2001-06-12 International Business Machines Corporation Relocating unreliable disk sectors when encountering disk drive read errors with notification to user when data is bad
US6332204B1 (en) 1999-03-31 2001-12-18 International Business Machines Corporation Recovering and relocating unreliable disk sectors when encountering disk drive read errors
US6426928B1 (en) 1999-03-31 2002-07-30 International Business Machines Corporation Ability to distinguish true disk write errors
US6941488B2 (en) * 2000-08-04 2005-09-06 Seagate Technology Llc Retrieval of a single complete copy from multiple stored copies of information
US6948165B1 (en) * 2001-02-28 2005-09-20 Western Digital Ventures, Inc. Method for installing an application program, to be executed during each bootload of a computer system for presenting a user with content options prior to conventional system startup presentation, without requiring a user's participation to install the program
JP4037617B2 (en) * 2001-03-16 2008-01-23 株式会社東芝 Defect search method
US7236987B1 (en) 2003-02-28 2007-06-26 Sun Microsystems Inc. Systems and methods for providing a storage virtualization environment
US7383381B1 (en) 2003-02-28 2008-06-03 Sun Microsystems, Inc. Systems and methods for configuring a storage virtualization environment
US7290168B1 (en) 2003-02-28 2007-10-30 Sun Microsystems, Inc. Systems and methods for providing a multi-path network switch system
US7430568B1 (en) 2003-02-28 2008-09-30 Sun Microsystems, Inc. Systems and methods for providing snapshot capabilities in a storage virtualization environment
US7339873B2 (en) * 2003-07-07 2008-03-04 Sony Corporation Data recording/reproducing apparatus, data recording/reproducing method, program, and recording medium
US20050066230A1 (en) * 2003-09-23 2005-03-24 Bean Robert George Data reliabilty bit storage qualifier and logical unit metadata
JP2006252733A (en) * 2005-03-14 2006-09-21 Fujitsu Ltd Medium storage device and write path diagnosing method for the same
US20060253674A1 (en) * 2005-05-06 2006-11-09 Xiv Ltd. Automatic disk healing
JP4038216B2 (en) * 2005-05-10 2008-01-23 ファナック株式会社 Sequence program editing device
US7831882B2 (en) * 2005-06-03 2010-11-09 Rambus Inc. Memory system with error detection and retry modes of operation
KR100827227B1 (en) * 2005-06-24 2008-05-07 삼성전자주식회사 Method and apparatus for managing DRM right object in low-processing power's storage efficiently
US7788555B2 (en) * 2005-07-22 2010-08-31 Broadcom Corporation Using fractional sectors for mapping defects in disk drives
US7480829B2 (en) * 2005-11-15 2009-01-20 International Business Machines Corporation Method, system and computer program product for recovery of formatting in repair of bad sectors in flash memory
CN109445696B (en) * 2018-10-21 2021-10-08 山西达鑫核科技有限公司 Information storage equipment for network aggregation storage

Family Cites Families (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3059266A (en) 1961-04-28 1962-10-23 Ibm Magnetic record processing apparatus
US3689891A (en) * 1970-11-02 1972-09-05 Texas Instruments Inc Memory system
JPS474054U (en) * 1971-01-28 1972-09-08
US3787815A (en) * 1971-06-24 1974-01-22 Honeywell Inf Systems Apparatus for the detection and correction of errors for a rotational storage device
US3771143A (en) * 1972-06-01 1973-11-06 Burroughs Corp Method and apparatus for providing alternate storage areas on a magnetic disk pack
US3997876A (en) * 1972-06-07 1976-12-14 International Business Machines Corporation Apparatus and method for avoiding defects in the recording medium within a peripheral storage system
US4037091A (en) 1976-04-05 1977-07-19 Bell Telephone Laboratories, Incorporated Error correction circuit utilizing multiple parity bits
FR2379112A1 (en) * 1977-01-27 1978-08-25 Cii Honeywell Bull METHOD OF WRITING INFORMATION CONCERNING DEFECTS OF A MAGNETIC RECORDING MEDIA
FR2426938A1 (en) * 1978-05-26 1979-12-21 Cii Honeywell Bull DEVICE FOR DETECTION OF DEFECTIVE SECTORS AND ALLOCATION OF REPLACEMENT SECTORS IN A DISK MEMORY
US4214280A (en) 1978-05-30 1980-07-22 Xerox Corporation Method and apparatus for recording data without recording on defective areas of a data recording medium
USRE31069E (en) 1978-09-11 1982-10-26 International Business Machines Corporation Apparatus and method for record reorientation following error detection in a data storage subsystem
JPS5674806A (en) * 1979-11-21 1981-06-20 Toshiba Corp Detecting process system for defective track
NL8004598A (en) * 1980-08-14 1982-03-16 Philips Nv METHOD FOR REGISTRATION IN, REPECTIVE READING FROM, A REGISTRATION BODY, SECTOR-ORGANIZED INFORMATION, AND DEVICE FOR IT.
JPS57105814A (en) * 1980-12-22 1982-07-01 Fujitsu Ltd Error correction processing system for magnetic disc device
JPS57143705A (en) * 1981-03-03 1982-09-06 Nec Corp Faulty sector processing method of magnetic disk medium
JPS5877034A (en) * 1981-10-30 1983-05-10 Hitachi Ltd Controlling system for unrewritable storage device

Also Published As

Publication number Publication date
US4434487A (en) 1984-02-28
JPS62162284A (en) 1987-07-18
JPS62162283A (en) 1987-07-18
JPS62162285A (en) 1987-07-18
EP0432137A2 (en) 1991-06-12
JPS62164278A (en) 1987-07-20
EP0090040A4 (en) 1986-01-28
DE3280359D1 (en) 1991-10-31
JPH0810535B2 (en) 1996-01-31
EP0090040B1 (en) 1991-09-25
EP0430928A2 (en) 1991-06-05
JPH02230559A (en) 1990-09-12
EP0430928A3 (en) 1992-09-02
JPS58501646A (en) 1983-09-29
EP0432137A3 (en) 1992-08-26
WO1983001334A1 (en) 1983-04-14
EP0090040A1 (en) 1983-10-05

Similar Documents

Publication Publication Date Title
CA1187615A (en) Disk format for secondary storage system
US4656532A (en) Sector identification method for hard sectored hard files
US6384999B1 (en) Rewrite with embedded reassign for data recovery from marginally defective data sites on a data storage device
US6263459B1 (en) On the fly write reallocation
US6181497B1 (en) System and method for providing nonadjacent redundancy synchronization bytes
JP2776006B2 (en) Information recording / reproducing device
US5463758A (en) System and method for reducing seek time for read operations in mirrored DASD files
EP0428208B1 (en) Sector slip with address collision recovery for write once recording media
US4914529A (en) Data disk defect handling using relocation ID fields
CN101071611B (en) Information recording /reproducing apparatus and recording/reproducing method
US6574699B1 (en) Fast track reassign in a rotating storage media
EP0428108A2 (en) Building emergency exhaust fan system
EP0282036B1 (en) Apparatus for recording data into optical recording medium
EP0429435A2 (en) Disk format for secondary storage system
US6141176A (en) Data storage disk with information encoded in the position error signal fields
JPH06314174A (en) Information recording medium and information recording and reproducing device
CA1205193A (en) Disk format for secondary storage system
EP0437202B1 (en) Information recording/reproducing apparatus for recording information and verifying recorded information
EP0435852A2 (en) Disk format for secondary storage system
EP0435853A2 (en) Disk format for secondary storage system
JP3993921B2 (en) Storage device
WO2001031444A2 (en) Method for preventing repeating non-recoverable read errors at same physical location on data storage media
EP0474052B1 (en) Zone access method in an optical disk drive apparatus

Legal Events

Date Code Title Description
MKEC Expiry (correction)
MKEX Expiry