JP4869416B2 - Magnetic disk device and data refresh method in the same - Google Patents

Description

  The present invention relates to a magnetic disk apparatus suitable for refreshing data written on a disk and a data refresh method in the apparatus.

  In recent years, the storage capacity per surface (recording surface) of a disk (magnetic disk) mounted on a magnetic disk device has been increased. In order to increase the storage capacity per surface of the disk, the linear recording density and / or the track density may be increased. In order to increase the linear recording density, the number of sectors per track may be increased. In order to increase the track density, that is, in order to secure more tracks in the plane of the disk, the distance between adjacent tracks (that is, the track pitch) may be shortened by narrowing the track width.

  Here, it is assumed that the track pitch is shortened in order to increase the track density. In general, the track pitch can be shortened only to such an extent that the data of the track (adjacent track) adjacent to the data write destination is not deteriorated by the data writing by the head (magnetic head). Each track has the same width as the width of the head (the write element included in the head). However, the width of the distribution of the magnetic field (recording magnetic field) generated by the head at the time of data writing does not necessarily match the width of the head, and the magnetic field is applied (leaks) to the periphery of the head. For this reason, if the track pitch is made too short, the degree to which the leakage magnetic field from the head affects the adjacent track becomes high when data is written to the track on the disk by the head.

  However, even when the track pitch is shortened beyond the limit, if the degree is small, the data in the adjacent track is not immediately destroyed by the data write operation once to several degrees. That is, generally, the data of the adjacent track gradually deteriorates according to the number of times by repeating the data writing operation.

  Therefore, in recent magnetic disk devices, data refresh (rewriting) for recovering data deterioration of adjacent tracks due to data write operation is becoming essential. For example, Patent Document 1 describes a technique (prior art) for refreshing data of a track adjacent to a track whose number of data writes has reached a specified number.

  According to the above prior art, the data of the track adjacent to the track whose number of data write has reached the specified number is read and temporarily stored in the RAM. Then, the data of the adjacent track temporarily stored in the RAM is rewritten to the adjacent track. That is, the data of the track to be refreshed is rewritten with the data read from the track. Data deterioration is recovered by such data rewriting processing, that is, data refresh processing. Therefore, by applying the data refresh process, the track pitch can be shortened beyond the conventional limit.

JP 2004-273060 A

  However, in the above prior art, there is a possibility that data to be refreshed may be lost if the power of the magnetic disk device is cut off during the data refresh process. More specifically, a magnetic disk device during an operation (data write operation) for rewriting data read from a track to be refreshed (hereinafter referred to as a target track) to the track. If the power supply of the PC is cut off, the data may be lost. The reason for this will be described.

  First, it is assumed that the power supply is cut off during the data write operation to the target track again. In this case, since the data writing operation to the target track is not completed, the data on the track is destroyed. At this time, the data of the target track temporarily stored in the RAM is lost. For this reason, even if the power is restored, the data write operation which has not been completed cannot be performed again, and the data of the target track is lost.

  In order to prevent such a problem, it is conceivable to use a specific track on the disk instead of the RAM. In other words, it is conceivable to use a specific track on the disk as a temporary saving track for temporarily saving the data of the target track. In this case, the data read from the target track is written (temporarily saved) on the temporary save track on the disk. Next, the read data is rewritten on the target track (that is, the track on which the data has been written). Even if the power is cut off during the data writing to the target track again, the data written (saved) on the temporary saving track is not lost. For this reason, the data writing operation which has not been completed can be performed again by using the data written in the temporary saving track.

  However, in order to refresh the data of the target track by using a temporary saving track on the disk instead of the RAM, the data read from the target track is rewritten before the data read from the target track is rewritten. A write operation for writing (temporarily saving) to the temporary saving track is required. Since this writing operation takes a long time compared to the case of writing data to the RAM, the efficiency of the data refresh process is lowered. Furthermore, a seek operation for moving the head from the target track to the temporary retreat track and a seek operation for moving the head from the temporary retreat track to the target track are also required. This two seek operations further reduce the efficiency of the data refresh process.

  As described above, by using a temporary saving track on the disk instead of the RAM, it is possible to cope with the power interruption, but there is a possibility that the efficiency of the data refreshing process is lowered.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a magnetic disk device and a data refresh method in the device capable of preventing the efficiency of data refresh processing from being lowered while coping with power interruption. It is in.

According to one aspect of the present invention, a magnetic disk device that performs data refresh is provided. The magnetic disk drive, a magnetic disk Lud over data track group comprising a preliminary data tracks and a plurality of data tracks that are used in the data refreshing is plurality, any data track group of the plurality of data tracks Group when refreshing the plurality of data tracks of the inner, data recorded in the data track adjacent to the front Ki予 Bei data track in the data track group is moved prior to the Ki予 Bei data track, according to this It said adjacent data tracks Te are repeated switching of the next preliminary data tracks, preliminary data track in the data track group; and a controller for controlling the so that is moved in order.

  According to the present invention, since the operation of writing the data of the track to be refreshed to the disk (temporary saving track) for temporary saving becomes unnecessary, it is possible to transfer the data to the disk as compared with the case of using the temporary saving track. The number of data write operations can be reduced once, and the efficiency of the refresh process can be prevented from being lowered while coping with power interruption.

1 is a block diagram showing a configuration of a magnetic disk device (HDD) according to an embodiment of the present invention. FIG. 3 is a conceptual diagram showing a format including a track arrangement of the disc shown in FIG. The conceptual diagram of the data refresh process applied in the embodiment. The figure which shows the data structure example of the write count table applied in the embodiment. The figure which shows the data structure example of the track shift table applied in the embodiment. 2 is a flowchart for explaining an overall operation in the HDD of FIG. 1. The flowchart which shows the detailed procedure of the calculation process of the real cylinder number contained in the flowchart of FIG. 7 is a flowchart showing a detailed procedure of track refresh processing included in the flowchart of FIG. 6. The conceptual diagram about optimization of the unit of refresh for the purpose of shortening the rotation waiting time by the refresh process of a reverse direction.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a magnetic disk device (HDD) 100 according to an embodiment of the present invention. An HDD 100 shown in FIG. 1 is a storage device for writing data on a recording surface of a disk (magnetic disk) 101 or reading data from the recording surface in response to a request from the host system 200. The host system 200 is an electronic device such as a personal computer that uses the HDD 100 as a storage device.

  The disk 101 is fixed to a spindle motor (SPM) 3 and rotates at a constant speed when the SPM 103 is driven. For example, one disk surface of the disk 101 forms a recording surface on which data is magnetically recorded. A head (magnetic head) 102 is arranged corresponding to the recording surface of the disk 101. The head 102 is fixed to one end of an actuator 105, and the other end of the actuator 105 is fixed to a voice coil motor (VCM) 104. When the VCM 104 is driven, the head 102 moves in a range that overlaps the surface of the disk 101 in an arc trajectory centered on the axis of the VCM 104.

  In the configuration of FIG. 1, an HDD 100 including a single disk 101 is assumed. However, the configuration may be such that the plurality of disks 101 are fixed to the SPM 103 with a certain gap. In this case, the plurality of actuators 105 are fixed to the VCM 104 in a state where they overlap with the gaps of the plurality of disks 101. A head 102 is fixed to one end of each of the plurality of actuators 105. Therefore, when the SPM 103 is driven, all the disks 101 rotate simultaneously, and when the VCM 104 is driven, all the heads 102 move simultaneously. In the configuration of FIG. 1, one surface of the disk 101 forms a recording surface. However, both sides of the disk 101 may form recording surfaces, and the heads 102 may be arranged corresponding to both recording surfaces.

FIG. 2 is a conceptual diagram showing a format including the track arrangement of the disk 101.
In FIG. 2, a plurality of tracks 201 are concentrically arranged on the recording surface of the disk 101. Data received by the HDD 100 from the host system 200 is recorded on at least one of the plurality of tracks 201 in accordance with an address designated by the host system 200.

  Servo areas 202 and data areas 203 are alternately arranged at equal intervals on the plurality of tracks 201 of the disk 101. Servo signals used for positioning the head 102 are recorded in the servo area 202. The data area 203 is used for recording data transferred from the host system 200.

  Referring again to FIG. 1, the CPU 115 functions as a main controller of the HDD 100. The CPU 115 performs control for starting / stopping the SPM 103 and maintaining the rotation speed via the motor driver 106. The CPU 115 also controls to drive the VCM 104 via the motor driver 106 to move the head 102 to the target track and to set the head within the target range of the track. Control for moving the head 102 to the target track is called seek control, and control for setting the head 102 within the target range of the target track is called head positioning control. The CPU 115 further performs control (data refresh processing) for refreshing data written on the track 201 of the disk 101.

  Positioning of the head 102 is performed in a steady rotation state after the start of the SPM 103. As described above, the servo areas 202 (see FIG. 2) are arranged at equal intervals in the circumferential direction of the disk 101. For this reason, servo signals recorded in the servo area 202 appear at regular intervals in the analog signal read from the disk 101 by the head 102 and amplified by the head IC 107. The read / write IC 108 (servo block 121 included in the read / write IC 108) and the gate array 109 generate the signal for positioning the head 102 by processing the analog signal using this. To do. The CPU 115 controls the motor driver 106 based on this signal, so that the current (VCM current) for positioning the head 102 is supplied from the motor driver 106 to the VCM 104 in real time.

  The CPU 115 controls the entire HDD, such as control of other ICs and command processing, while controlling the SPM 103 and VCM 104 via the motor driver 106 as described above. The CPU 115 is connected to the CPU bus 112.

  A read / write IC 108, a gate array 109, a disk controller (HDC) 110, a RAM 113, and a flash ROM 114 are connected to the CPU bus 112. The flash ROM 114 is a rewritable nonvolatile memory. Here, rewriting of the flash ROM 114 is performed under the control of the CPU 115. A program to be executed by the CPU 115 is stored in the flash ROM 114 in advance. The above-described control by the CPU 115 is realized by the CPU 115 executing the program. A part of the storage area of the flash ROM 114 is used to store a track shift table 500 (see FIG. 5) described later. The RAM 113 is used for storing various variables used by the CPU 115, for example. A part of the storage area of the RAM 113 is used as a work area of the CPU 115.

  The read / write IC 108 includes a servo block 121 and a read / write block 122. The servo block 121 performs signal processing necessary for positioning of the head 102 including extraction of servo signals. The read / write block 122 performs signal processing for reading / writing data. The gate array 109 generates various control signals including a signal for extracting a servo signal by the servo block 121.

  In addition to the CPU bus 112, the HDC 110 is connected to a read / write IC 108, a gate array 109, and a buffer RAM 111. The HDC 110 includes a host block 123, a read / write block 124, and a buffer block 125. The host block 123 has a host interface control function for receiving a command (write command, read command, etc.) transferred from the host system 200 and controlling data transfer between the host and the HDC 110. The read / write block 124 is connected to the read / write IC 108 and the gate array 109, and performs read / write processing of data performed via the read / write IC 108. The buffer block 125 controls the buffer RAM 111. A part of the storage area of the buffer RAM 111 is used as a write buffer for temporarily storing data (write data) to be written to the disk 101 via the HDC 110 (read / write block 124 in the HDC 110). Another part of the storage area of the buffer RAM 111 is used as a read buffer for temporarily storing data (read data) read from the disk 101 via the HDC 110.

  The read / write IC 108, the gate array 109, and the HDC 110 each have a control register. Each of these control registers is allocated to a part of the memory space of the CPU 115, and the CPU 115 accesses this part of the area, so that the read / write IC 108 and the gate array 109 are accessed via the control register. Or the HDC 110 is controlled.

  In the HDD 100 of FIG. 1, data is read as follows. First, a signal (analog signal) read from the disk 101 by the head 102 is amplified by the head IC 107. The amplified analog signal is separated into a servo signal and a data signal by the read / write IC 108. The data signal is decoded by the read / write block 122 in the read / write IC 108 and then sent to the HDC 110. The read / write block 124 in the HDC 110 processes the decoded data signal according to a control signal from the gate array 109 to generate data to be transferred to the host system 200. The processing here includes detection of data errors based on ECC data, which will be described later, and correction of errors. The generated data is temporarily stored in the buffer RAM 111 by the buffer block 125 in the HDC 110 and then transferred to the host system 200 by the host block 123 in the HDC 110.

  In the HDD 100 of FIG. 1, data is written as follows. The data transferred from the host system 200 to the HDC 110 is received by the host block 123 in the HDC 110 and then temporarily stored in the buffer RAM 111 by the buffer block 125 in the HDC 110. The data stored in the buffer RAM 111 is taken out by the buffer block 125 and then sent to the read / write IC 108 by the read / write block 124 in the HDC 110 in accordance with a control signal from the gate array 109. The data sent to the read / write IC 108 is encoded by the read / write block 122 in the read / write IC 108. The encoded data is sent to the head 102 via the head IC 107 and written on the disk 101 by the head 102. The above-described data reading / writing is performed under the control of the CPU 115.

Next, an outline of data refresh processing executed in the HDD 100 of FIG. 1 will be described.
As described in the section of the problem to be solved by the invention, as a countermeasure against power interruption during data refresh, the data of the track (target track) to be refreshed is stored on a specific track (for temporary saving) on the disk. If the data is temporarily saved in the (track), the efficiency of the data refresh process is lowered. Therefore, in the present embodiment, the following characteristic data refresh processing that can prevent a decrease in processing efficiency while applying power cut-off measures during data refresh is applied.

  In this embodiment, a spare track Ti is prepared on the disk 101. A track Tj adjacent to the spare track Ti becomes a track (target track) Tj to be subjected to data refresh.

  First, data is read from the target track Tj by the head 102. Next, the head 102 is moved from the target track Tj to the spare track Ti. Here, since the spare track Ti is adjacent to the target track Tj, the head 102 can be moved from the target track Tj to the spare track Ti at high speed.

  When the head 102 is moved to the spare track Ti, the data read from the target track Tj is written to the spare track Ti by the head 102. Thereby, the refresh of the data of the target track Tj is completed. That is, the feature of the present embodiment is that the data of the target track Tj is written on the adjacent spare track Ti to refresh the data on the spare track Ti. Reading of data from the target track Tj and writing of the read data to the spare track Ti are performed under the control of the CPU 115.

  Even if the power is cut off during the writing operation for writing the data read from the target track Tj to the spare track Ti, the data of the target track Tj is not lost. Therefore, when the power is restored, the data can be read from the target track Tj and the read data can be written to the spare track Ti.

  Thus, in this embodiment, it should be noted that the data read from the target track Tj is not rewritten to the target track (that is, the original track) Tj via the temporary saving track. is there. According to this embodiment, the operation of writing the data of the target track to the disk 101 (temporary saving track) for temporary saving becomes unnecessary. That is, according to the present embodiment, the data write operation to the disk 101 can be reduced once compared with the case where the data of the target track is refreshed using the temporary save track, and the refresh is performed while coping with the power shutdown. It can prevent that the efficiency of processing falls.

  When the data read from the target track Tj is written to the spare track Ti, the target track Tj becomes a new spare track Tj. Then, the track Tk adjacent to the new spare track Tj on the side opposite to the track Ti becomes the new target track Tk. Then, the head 102 is moved from the track Ti to a new target track Tk. Since the distance between the track Ti and the new target track Tk is two tracks, the head 102 can be moved from the track Ti to the new target track Tk at high speed.

  Next, data is read from the target track Tk by the head 102. Next, the head 102 is moved from the target track Tk to the adjacent new (current) spare track Tj. Then, the data read from the target track Tk is written by the head 102 to the current spare track Tj. Similarly, the spare tracks are sequentially switched in units of one track, and each time, the operation of writing the data of the track adjacent to the spare track at that time to the spare track at that time is performed.

  According to this embodiment, since the target track and the spare track are always adjacent to each other, the head 102 can be moved from the target track to the spare track at a high speed. When the data of the target track is refreshed on the spare track, the target track becomes a new spare track. Then, a track adjacent to the new spare track on the opposite side to the track that has been used as a spare track (the old spare track) becomes a new target track. Since the distance between the old spare track and the new target is two tracks, the head 102 can be moved from the old spare track to the new target track at high speed. Therefore, according to the present embodiment, the time required for moving the head can be shortened as compared with the case where the data of the target track is refreshed using the temporary saving track. Also from this point, according to the present embodiment, it is possible to prevent the efficiency of the refresh process from being lowered while coping with the power interruption.

Next, the data refresh process described above will be described with reference to the conceptual diagram of FIG.
In the present embodiment, the set of tracks 201 arranged on the recording surface of the disk 101 is grouped in units of a predetermined number of tracks, and data refresh processing is performed in units of groups. This group is called a track group. Here, for each track group, the total number of data write operations (the number of writes) to the entire track group is counted. When the count value reaches a predetermined constant value, data refresh processing for the corresponding track group is performed.

  In the example of FIG. 3, one track group on the disk 101 is shown. Here, for simplification of explanation, it is assumed that the track group is composed of 11 tracks of tracks 0 to 10. The numerical values “0” to “10” in the notation of tracks 0 to 10 indicate the relative positions (relative track positions) of the tracks in the track group. It is assumed that a track having a smaller value is positioned on the outer peripheral side on the disk 101, and a track having a larger value is positioned on the inner peripheral side on the disk 101.

  In FIG. 3, a state 31 indicates a state where one spare track is secured in one track group. Here, the track 10 is assigned to the spare track. However, the spare track is dynamically switched as will be described later. It is assumed that data A to J are stored in tracks 0 to 9 other than the spare track 10 in the track group, respectively.

  Assume that the track group in the state 31 is the target of data refresh. In this case, in the present embodiment, the track 9 adjacent to the spare track 10 (on the outer periphery side of the disk 101) is first selected as a data refresh target track (target track). Then, the data J of the track 9 is read, and the read data J is written to the track 10 which is a spare track at that time, as indicated by an arrow 311 in FIG. As a result, the data of the track 9 is refreshed on the track 10 adjacent to the track 9.

  Then, the track 9 that has been subject to data refresh until then is newly assigned to the spare track in place of the track 10. In this state, the track 8 adjacent to the new spare track 9 (on the outer peripheral side of the disk 101) is selected as the target track. Then, the data I of the track 8 is read, and the read data I is written in the track 9 which is a spare track at that time, as indicated by an arrow 312 in FIG. As a result, the data of the track 8 is refreshed on the track 9 adjacent to the track 8.

  Thereafter, the same operation is sequentially performed while switching the spare track. Eventually, data B of track 1 is written to track 2 as shown by arrow 319 in FIG. 3, and data A of track 0 is written to track 1 as shown by arrow 320 in FIG. It is assumed that 0 is newly assigned to the spare track. Thus, when the track 0 becomes a spare track, that is, when the track group in the state 31 in FIG. 3 transitions to the state 32 shown in FIG. 3, a data refresh process (data refresh process) for the track group is performed. Operation) is completed. This is called a forward refresh process. That is, the data refresh process in which the data of the target track is written in a track (preliminary track) adjacent to the inner circumferential direction of the disk 101 (the direction in which the value of the relative track position increases) is called a forward refresh process. In addition, this direction is called a refresh direction, and a write operation for writing data of a target track to an adjacent track is called shift writing (shift write).

  Thereafter, it is assumed that the data refresh process for the track group in the state 32 in FIG. 3 is required. In this case, contrary to the forward refresh process, the track 1 adjacent to the spare track 0 is selected as the target track. Then, the data A of the track 1 is read, and the read data A is written to the track 0 which is a spare track at that time, as indicated by an arrow 321 in FIG. As a result, the data A of the track 1 is refreshed on the track 0 adjacent to the track 1.

  Then, the track 1 that has been the object of data refresh until then is newly assigned to the spare track in place of the track 0. In this state, the track 2 adjacent to the new spare track 1 (on the inner periphery side of the disk 101) is selected as the target track. Then, the data B of the track 2 is read, and the read data B is written to the track 1 which is a spare track at that time, as indicated by an arrow 322 in FIG. As a result, the data B of the track 2 is refreshed on the track 1 adjacent to the track 2.

  Thereafter, the same operation is sequentially performed while switching the spare track. Eventually, data I of track 9 is written to track 8 as shown by arrow 329 in FIG. 3, and data J of track 10 is written to track 9 as shown by arrow 330 in FIG. Assume that 10 is newly assigned to the spare track. In this way, when the track 10 becomes a spare track, that is, when the track group in the state 32 in FIG. 3 transitions to the state 33 shown in FIG. 3, the data refresh process for the track group is completed. . This is called a reverse refresh process. In other words, the refresh process in which the data of the target track is written in a track (preliminary track) adjacent to the outer peripheral direction of the disk 101 (the direction in which the value of the relative track position decreases) is called a reverse refresh process.

  Thereafter, the refresh process is performed while switching the refresh direction every time the track group shown in FIG. 3 needs to be refreshed. That is, the forward refresh process and the reverse refresh process are alternately performed. In the example of FIG. 3, the number of tracks per track group is 11 including the spare tracks for the sake of simplicity. In this example, the actual storage capacity of the disk 101 can be secured only 10/11 of the total number of tracks (cylinders). Therefore, in practice, it is preferable to set a larger value as the number of tracks per track group.

  FIG. 4 shows an example of the data structure of the write count table 400 that holds the number of writes (data write count, write execution count) for each track group. The write count table 400 is stored, for example, in a predetermined area of the RAM 113 in FIG.

  In the example of the write count table 400 in FIG. 4, for the sake of general description, it is assumed that the HDD 100 includes m heads 102 and the HDD 100 includes n cylinder groups. In this case, the write count table 400 shows the number of times of writing to each track group (the number of data writes) for each of the track groups represented using the head (head number) h and the cylinder group (cylinder group number) c. W (h, c) (0 ≦ h ≦ m−1, 0 ≦ c ≦ n−1) is held. W (h, c) is used as a write counter that counts the number of writes to the track group specified by the head number h and the cylinder group number c. Note that m is 1 in the case of the configuration of the HDD 100 of FIG.

  A cylinder group is a set of a predetermined number of cylinders, and the number of cylinders per cylinder group is the same as the number of tracks per track group. Therefore, there are as many track groups having the same cylinder group number in the HDD 100 as the number m of the heads 102. A track group is specified by a cylinder group number c and a head number h. When data is written (write access) to a track in the track group specified by the cylinder group number c and the head number h, the number of writes (write counter) W (h, c) is incremented by the number of times writing has been performed.

  In the present embodiment, the write count table 400 is stored in the RAM 113 as described above. The contents of the RAM 113 are lost when the HDD 100 is powered off. Therefore, the contents of the write count table 400 are lost when the power is turned off. For this reason, in the present embodiment, the contents of a predetermined area in the RAM 113 including the write count table 400 are stored in a predetermined area of the disk 101 as needed (for example, when the HDD 100 shifts to the power saving state). (Evacuated). The contents including the count table 400 stored in a predetermined area of the disk 101 are read when the HDD 100 is activated (when power is turned on) and restored in the RAM 113.

  In the present embodiment, the track in which the target data is written is switched every time data refresh is performed on the track group including the track. For example, in FIG. 3, the track in which data A is written is track 0 in state 31 or 33, and is track 1 in state 32. In order to access the target data, it is always necessary to keep track of the information indicating in what direction and how far the above-described shift writing has been performed, and to correct the cylinder number deviation with respect to the result of normal address conversion. .

  Here, normal address translation will be described. In recent HDDs, a so-called CDR (Constant Density Recording) system is generally applied. It is assumed that the CDR system is also applied to the HDD 100 in FIG. Generally, the host system 200 using such an HDD 100 gives an LBA (Logical Block Address) to the HDD 100 as information for designating an access target block for the HDD 100. In this case, in the HDD 100, for example, when executing a command such as a read command or a write command instructed from the host system 200, the LBA (logical block address) given by the host system 200 is physically stored on the disk 101. Address conversion (address calculation) is performed to convert the data into cylinder numbers, head numbers, and sector numbers (physical addresses consisting of the positions). This is normal address translation.

  In the present embodiment, position information indicating the position (relative position) of the spare track in the track group is provided for each track group so that the deviation of the cylinder number can be corrected with respect to the result of such normal address conversion. Retained and managed. In addition, in order to correct the deviation of the cylinder number, it is necessary to grasp the direction of the shift writing (refresh direction). However, the refresh direction can be determined from the position of the spare track in the track group as will be described later. For this reason, in this embodiment, a configuration in which information indicating the refresh direction is held and managed for each track group is not applied. However, for the track group in which the refresh process is being executed (or interrupted), the refresh direction of the refresh process is managed by the refresh management information described later.

  FIG. 5 shows an example of the data structure of the track shift table 500 that holds the position of the spare track for each track group. In the example of the track shift table 500 in FIG. 5, for the sake of generalization, it is assumed that the HDD 100 includes m heads 102 and the HDD 100 includes n cylinder groups. In this case, the track shift table 500 includes, for each of all the track groups represented using the head (head number) h and the cylinder group (cylinder group number) c, the position S (h) of the spare track in the track group. , C) (0 ≦ h ≦ m−1, 0 ≦ c ≦ n−1). The spare track position S (h, c) indicates the relative position (relative track position) of the spare track in the track group to which the spare track belongs. Note that m is 1 in the case of the configuration of the HDD 100 of FIG.

  The track shift table 500 is preferably stored in a rewritable nonvolatile memory as a countermeasure against power supply means. The data amount of the track shift table 500 is small. Therefore, in this embodiment, a configuration in which the track shift table 500 is stored in a predetermined area of the flash ROM 114 used for storing the program is applied.

  As described above, in the HDD 100, normal address conversion for converting the LBA given by the host system 200 into the cylinder number, the head number, and the sector number indicating the physical position on the disk 101 is performed. However, in the present embodiment, the cylinder number of the target track changes depending on the position of the spare track in the track group and the refresh direction. Therefore, in addition to normal address conversion, it is necessary to correct the cylinder number according to the position of the spare track and the refresh direction indicated by the track shift table 500.

  Thus, the contents of the track shift table 500 are extremely important for the correction process for determining the cylinder number of the target track. For this reason, the track shift table 500 should never be lost even if an unexpected power interruption occurs during the operation of the HDD 100. Therefore, the track shift table 500 is stored in the flash ROM 114.

  On the other hand, in contrast to the track shift table 500, the write count table 400 is stored in the RAM 113. For this reason, when the power supply is unexpectedly shut down during the operation of the HDD 100, after the HDD 100 is restarted, the content of the write count table 400 last stored in the disk 101 before that is used. In this case, the lost data is almost unaffected because it is only the number of writes from the last saving of the write count table 400 to the power shutdown.

  On the other hand, in the track shift table 500, if it is assumed that the update of the spare track position S (h, c) in a certain track group for a certain period is lost, the cylinder numbers of the tracks in the track group cannot be corrected. . For this reason, the track shift table 500 must always be stored in a complete form. Therefore, if it is assumed that the track shift table 500 is stored in the RAM 113 in the same manner as the write count table 400, the contents of the track shift table 500 are saved in a predetermined area of the disk 101 every time the data of one track is refreshed. (Evacuation) must be done. In this way, the efficiency of the refresh process cannot be achieved.

  Therefore, in the present embodiment, a configuration in which the track shift table 500 is stored in the flash ROM 114 as described above is applied. Here, it is preferable to prepare two primary and secondary track shift tables 500 so as to cope with power interruption during updating of the track shift table 500. That is, these two primary and secondary tables (track shift table) 500 are stored in different pages of the flash ROM 114, and the table is updated with respect to the two primary and secondary tables. Here, a configuration is applied in which the update of the secondary table 500 is executed after the update of the primary table 500 is normally completed. As a result, the contents of the track shift table 500 can be prevented from being lost no matter what timing the HDD 100 is powered off.

  In general, the flash ROM 114 is rewritten by a procedure of, for example, erasing all the contents in units of pages and then writing new contents on the erased pages. However, depending on the type of the flash ROM 114, it is possible to add information on the page. If the flash ROM 114 is a type that can be additionally written in a page, a difference area is provided in the flash ROM 114, and the flash ROM 114 can be used as follows.

  For example, until the difference area in the flash ROM 114 is used up, instead of rewriting the entire track shift table 500, only the difference, that is, the changed S (h, c) is sequentially recorded in the difference area. Then, when there is no more space in the difference area, the entire track shift table 500 is updated. By applying such a configuration, the frequency of erasing the flash ROM 114 can be reduced. However, the track shift table 500 cannot be directly referred to by the method of recording the difference. Therefore, a track shift table 500 indicating the latest state may be arranged in a predetermined area of the RAM 113 and the track shift table 500 in the predetermined area may be referred to.

Next, the overall operation of the HDD 100 of FIG. 1 in which the refresh process for each track group is performed will be described with reference to the flowchart of FIG.
First, it is assumed that the power of the HDD 100 is turned on and the operation of the CPU 115 is started (step 601). Then, the CPU 115 performs known initialization processing and activation processing for the entire HDD 100 (step 602). When the startup process is completed, the CPU 115 is ready to receive a command from the host system 200 via the HDC 110, and enters a command wait loop (steps 603 to 607).

  When the reception of the command from the host system 200 is confirmed in step 603, the CPU 115 branches to step 611 to escape from the command wait loop and executes processing according to the command from the host system 200. In step 611, the CPU 115 determines whether the command from the host system 200 is a write command. If so, the CPU 115 performs processing for the write command (steps 612 to 615).

  If the received command is a write command (YES in step 611), in order to write to the address requested by the host system 200, the cylinder number correction as described above is necessary. In step 612, the cylinder number is corrected (calculation of the actual cylinder number). Although not explicitly shown in the flowchart of FIG. 6, before reaching step 612, an address (for example, LBA) specified by a command (write command) from the host system 200 is set as a virtual cylinder number and a head number. It is assumed that the process for converting to a sector number (normal address conversion) has already been performed by a general method.

  The virtual cylinder number is a calculated cylinder number provided in the middle to obtain an actual (normal) cylinder number (actual cylinder number). Here, the actual cylinder number is referred to as the actual cylinder number in comparison with the virtual cylinder number. A cylinder designated by a virtual cylinder number is called a virtual cylinder, and a cylinder designated by a real cylinder number is called a real cylinder. In addition, when it only expresses as a cylinder, it shall refer to an actual cylinder.

  The virtual cylinder number is a concept that is necessary when one spare track is secured for each track group (cylinder group), and tracks (cylinders) that can be recognized from the host system 200, excluding the spare track (spare cylinder). Indicates the cylinder number. That is, the virtual cylinder number is a cylinder number that is counted (calculated) by skipping a spare track in each track group (cylinder group) (without assuming the presence of the spare track).

  The concept of virtual cylinder number is introduced for the purpose of hierarchizing address translation (address calculation) as described below. First, the first address conversion (address calculation) is performed in the same manner as when it is assumed that there is no data refresh processing (track refresh processing) in units of track groups as applied in the present embodiment. In this first address conversion, the LBA specified by the host system 200 is converted into a cylinder number, a head number, and a sector number in the same manner as in normal address conversion. The spare track is not recognized from the host system 200. Therefore, the cylinder number acquired by the first address conversion from the LBA designated by the host system 200 is a virtual cylinder number. Therefore, this virtual cylinder number is converted into an actual cylinder number by the second address conversion (address calculation).

  As described above, in the present embodiment, when reaching step 612, the LBA designated by the host system 200 has already been converted into the virtual cylinder number, head number, and sector number. Therefore, in step 612, the CPU 115 performs conversion (calculation of the actual cylinder number) from the virtual cylinder number necessary for performing the write process to the actual cylinder number. Details of step 612 will be described later. When executing step 612, the CPU 115 performs write processing (control for data writing operation) based on the converted (calculated) actual cylinder number, the head number, and the sector number (step 613).

  When the write process (step 613) is completed, the CPU 115 updates the table 400 so that the write process is reflected in the write count table 400 (step 614). That is, when the write process is performed on the track group specified by the head h and the cylinder group c, the CPU 115 reflects the write count W (h, c) in the write count table 400 so that the write process is reflected. Update. More specifically, the CPU 115 adds the number of times of writing to the number of write times W (h, c) in the write count table 400. Here, usually, 1 is added. However, when a retry is performed in the write process, the retry affects the adjacent track in the same way as a normal write operation, so the number of retries is also added.

  When step 614 is executed, the processing of the write command is completed. Therefore, the CPU 115 performs command termination processing such as updating registers and releasing the busy state (step 615), and returns to the command wait loop.

  On the other hand, if the received command is other than a write command (NO in step 611), the CPU 115 performs processing (step 620) and command termination processing (step 615) according to the received command, and a command wait loop Return to. Processes corresponding to commands other than the write command are collectively described in step 620 for convenience. However, since there are actually various commands other than the write command, there are a number of commands corresponding to the determination of the command code as performed in step 611 and the processing corresponding to steps 612 and 613, and the number of commands. This is done in the same way as

  In particular, in a command such as a read command or a seek command in which the host system 200 designates an access target, a conversion process from a virtual cylinder number to a real cylinder number is performed in the same manner as in step 612 before command processing. There is a need.

  Next, assume that it is determined in step 603 in the command wait loop that no command has been received. In this case, idle time processing is performed. In addition, after the command termination process is performed in step 615, the idle process is performed. The idle time process includes a track refresh process. In this embodiment, the CPU 115 determines whether to perform the track refresh process before the track refresh process (steps 604 and 605).

  In step 604, the CPU 115 comprehensively determines whether it is necessary to immediately execute a command from the host system 200 without performing the track refresh process or whether the refresh process should be avoided. The command needs to be executed immediately when, for example, the command is received from the host system 200 immediately after step 615. The situation where the refresh process should be avoided is when the HDD 100 is subjected to vibration exceeding a certain level from the outside, or when the environmental temperature of the HDD 100 is out of the temperature range in which the operation of the HDD 100 is guaranteed. It is a case where it is used in. In step 605, the CPU 115 determines whether the track refresh process can be performed from the comprehensive determination result in step 604.

  Only when it is determined to perform the track refresh process, the CPU 115 performs the track refresh process (step 606). Details of step 606 will be described later.

  When the track refresh process at step 606 is completed or when it is determined at step 605 that the track refresh process should not be performed, the CPU 115 determines whether to execute the power saving process for shifting to the power saving state. If it is necessary to execute, the processing is executed (step 607). The power saving process includes a process such as an unload process for unloading the head 102 from the disk 101 and / or an SPM stop process for stopping the rotation of the SPM 103.

  When the power saving process is executed in step 607, the CPU 115 returns to step 603. On the other hand, if it is necessary to immediately execute a command from the host system 200, it is determined in step 607 that the power saving process should not be executed. In this case, the CPU 115 returns to step 603 without executing the power saving process. Thereafter, the CPU 115 repeats the above processing including step 603.

  Next, the detailed procedure of the actual cylinder number calculation process (cylinder number correction process) in step 612 will be described with reference to the flowchart of FIG.

  As described above, the virtual cylinder number applied in the present embodiment is a cylinder number that is counted by skipping the spare cylinder in each cylinder group, that is, a cylinder number that is counted as if there is no spare cylinder. For this reason, a virtual cylinder number cannot be directly converted into a real cylinder number. Therefore, in the actual cylinder number calculation process 701 shown in the flowchart of FIG. 7, the CPU 115 first determines, from the virtual cylinder number, the cylinder group number of the cylinder group to which the cylinder of the virtual cylinder number (virtual cylinder) belongs, and the virtual cylinder The offset (relative cylinder position) in the cylinder group is calculated (step 702).

  In the present embodiment, the number of virtual cylinders per cylinder group is a constant value, which is 10 in the example of FIG. The cylinder group number is the quotient (integer value) obtained by dividing the virtual cylinder number by this constant value, and the offset in the cylinder group is the remainder. For example, when the virtual cylinder number is 34, the cylinder group number (c) is 3 which is a quotient obtained by dividing 34 by 10, and the offset (relative cylinder position, relative virtual cylinder number) is 4 which is the remainder.

  Next, the CPU 115 uses the relative virtual cylinder number in the cylinder group indicated by the offset of the virtual cylinder calculated in step 702, that is, the cylinder group number calculated in step 702, as described below. Find the actual cylinder number.

  First, the CPU 115 refers to the track shift table 500 on the basis of the head number (h) obtained before reaching step 612 and the cylinder group number (c) calculated in step 702, so that a spare cylinder ( The position (relative cylinder position in the cylinder group) S (h, c) is obtained (step 703). As described above, when the cylinder group number (c) is 3 and the head number (h) is 1, the relative cylinder position in the cylinder group of the spare cylinder (track) is S (1, 3).

  The relative actual cylinder number is obtained by comparing the relative cylinder position S (h, c) (= S (1, 3)) within the cylinder group of the spare cylinder (track) acquired from the track shift table 500 and the offset within the cylinder group of the virtual cylinder ( (Relative virtual cylinder number). Specifically, the relative actual cylinder number is determined depending on whether the value indicating the relative cylinder position in the cylinder group of the reserve cylinder (track) is larger than the relative virtual cylinder number (offset) (step 704).

  When the value indicating the relative cylinder position in the cylinder group of the spare cylinder is larger than the relative virtual cylinder number (offset) (YES in step 704), the relative actual cylinder number (relative cylinder position in the cylinder group of the actual cylinder) is Not affected. For this reason, no deviation occurs between the relative virtual cylinder number and the relative actual cylinder number. That is, when the spare cylinder is present behind the actual cylinder (the cylinder number is larger), there is no deviation between the relative virtual cylinder number and the relative actual cylinder number. In this case, the relative real cylinder number is equal to the relative virtual cylinder number.

  On the other hand, when the value indicating the relative cylinder position in the cylinder group of the spare cylinder is equal to or less than the relative virtual cylinder number (offset) (NO in step 704), the relative actual cylinder number is affected by the spare cylinder. That is, when the spare cylinder does not exist behind the actual cylinder, the relative actual cylinder number is affected by the spare cylinder and is shifted by the amount of the spare cylinder. In this case, the relative real cylinder number is a value larger by 1 than the relative virtual cylinder number.

  Therefore, the CPU 115 further adds the actual cylinder number of the first cylinder of the cylinder group to the above-described relative actual cylinder number, that is, the product of the cylinder group number and the actual number of cylinders per cylinder group (the product of 3 and 11 in the above example). That is, the value obtained by adding 33) is acquired as the actual cylinder number (step 705 or 706).

  Here, step 705 is executed when step 704 is NO, and a value obtained by adding 1 to the relative virtual cylinder number (offset) acquired in step 702 is used as the relative actual cylinder number. In the above example, when S (1, 3) is equal to or less than the relative virtual cylinder number (offset) 4, step 705 is executed. By adding 4 and 1 to 33, 38 is obtained as the actual cylinder number.

  On the other hand, Step 706 is executed when Step 704 is YES, and the relative virtual cylinder number (offset) acquired in Step 702 is used as it is as the relative actual cylinder number. In the above example, when S (1, 3) is larger than the relative virtual cylinder number (offset) 4, step 706 is executed, and by adding 4 to 33, 37 is acquired as the actual cylinder number.

  Next, the detailed procedure of the track refresh process in step 606 will be described with reference to the flowchart of FIG.

  In the track refresh process 801 shown in the flowchart of FIG. 8, the CPU 115 determines whether there is a track group in which the refresh process is interrupted by referring to an execution flag described later (step 802). This step 802 relates to the fact that the refresh process can be interrupted for each track in the track group to be refreshed.

  The reason why the refresh process can be interrupted for each track is because it takes time to refresh the data of all the tracks (except the spare track) in the track group, and therefore the host system 200 received during the refresh process. This is because the responsiveness to the command from is not reduced. That is, in this embodiment, when a command from the host system 200 is received during execution of the refresh process, the refresh process is interrupted in order to prioritize the execution of the command.

  If there is a track group for which the refresh process is interrupted (YES in step 802), this means that the track group to be refreshed has been determined. In this case, the CPU 115 performs a refresh process after step 805.

  On the other hand, if there is no track group for which the refresh process is interrupted (NO in step 802), the CPU 115 searches the write count table 400 for a track group that can be the target of the refresh process (step 803). Here, the CPU 115 searches the write count table 400 for the write count W (h, c) having the largest value. Then, the CPU 115 determines whether or not there is a track group that requires refresh processing depending on whether or not the retrieved number of write times W (h, c) exceeds a predetermined value (step 804).

  When the retrieved number of write times W (h, c) exceeds a certain value, the CPU 115 has a track group that requires refresh processing, and associates the track group with the number of write times W (h, c). Is determined to be a track group (a track group represented by a head number h and a cylinder group number c) (YES in step 804). As described above, when it is determined that there is a track group that needs to be refreshed (that is, a track group to be refreshed), it is associated with the track group (that is, the write count W (h, c) having the largest value). Track group) is also identified. In this case, the CPU 115 performs a refresh process after step 805.

  On the other hand, when the retrieved number of write times W (h, c) is equal to or less than a certain value, the CPU 115 determines that there is no track group that requires refresh processing (NO in step 804). In this case, since it is not necessary to perform the refresh process, the CPU 115 returns from the track refresh process 801 to the original process (step 817). Thereby, step 607 in the flowchart of FIG. 6 is executed.

  By the way, the determination as to whether or not the track refresh process is to be executed is also performed in steps 604 and 605. However, in steps 604 and 605, as is clear from the above description, determination based on the write count table 400 as performed in steps 803 and 804, that is, whether there is a track group that needs to perform track refresh processing. No judgment is made.

  However, if it is necessary to refer to the write count table 400 in the steps 604 and 605, the same determination as in steps 803 and 804 may be performed in this step. For example, this may be the case where the determination of whether or not to perform the refresh process in steps 604 and 605 is performed based on the number of write times W (h, c) for each track group in addition to the environmental conditions such as vibration and temperature described above. It hits. Details of such determination are as follows. First, as the number of writes to a track group is small, data deterioration is generally small. For this reason, when the HDD 100 is used under adverse conditions, it is determined that the risk is lower when the refresh process is avoided than when the refresh process is performed. On the other hand, when the number of writes increases, data deterioration proceeds. For this reason, it is determined that the risk is lower when the refresh process is performed under some adverse conditions. If such a determination is made in steps 604 and 605, the track group with the largest number of writes is specified and the number of times is acquired in steps 604 and 605. In this case, these need not be performed again in steps 803 and 804.

  Now, at the stage where step 805 is executed, a track group (hereinafter referred to as a target track group) on which track refresh processing is to be performed has been determined. Therefore, in step 805, the CPU 115 refers to the track shift table 500 to obtain the relative position S (h, c) of the spare track in the target track group.

  When the process proceeds from step 802 to step 805 via steps 803 and 804, the CPU 115 performs refresh for the target track group based on the acquired relative position S (h, c) of the spare track in step 805. It is determined whether the processing direction (refresh direction) is the forward direction or the reverse direction. This determination method will be described later. In step 805, the CPU 115 further stores refresh management information for managing refresh in units of track groups in a predetermined area of the RAM 113.

  The refresh management information includes a track group number, a direction flag, and a running flag. The track group number indicates the track group (that is, the target track) that is specified in step 804 and that is to be refreshed. The direction flag indicates whether the direction (refresh direction) of the refresh process for the track group specified by the track group number is the forward direction or the reverse direction. Here, the direction flag indicates the refresh direction determined in step 805. The in-execution flag indicates whether the refresh process for the track group specified by the track group number is being executed (or interrupted). Here, the in-execution flag is set to indicate that the execution is in progress.

  On the other hand, when branching from step 802 to step 805, valid refresh management information is already stored in a predetermined area of the RAM 113 as described later. The in-execution flag included in the refresh management information is set to indicate that it is being executed (suspended). The direction flag included in the refresh management information indicates the direction of the refresh process being executed (suspended).

  The processing after step 805 is divided depending on whether the refresh processing for the target track group is performed in the forward direction or in the reverse direction.

  Therefore, the CPU 115 targets the track group (target track group) designated by the track group number included in the management information by referring to the direction flag included in the refresh management information stored in the RAM 113. It is determined whether or not the refresh process is performed in the forward direction (step 806).

  In the present embodiment, if there is a track group in which refresh is interrupted, the refresh process for the track group is preferentially performed. For this reason, a plurality of track groups are not simultaneously refreshed. Therefore, only one refresh management information exists for the entire HDD 100.

  As described above, the refresh management information is stored in the RAM 113. For this reason, the refresh management information is lost when the HDD 100 is powered off. Therefore, in the present embodiment, the refresh management information is stored in a predetermined area of the disk 101 when the HDD 100 is shifted to the power saving state, for example, as in the write count table 400. The refresh management information stored in a predetermined area of the disk 101 is read out by an initialization process that is executed when the HDD 100 is activated (at power-on), and is restored in the RAM 113.

  However, when an unexpected power interruption occurs, the latest refresh management information at that time is lost. Therefore, the CPU 115 determines whether the latest refresh management information has been lost, that is, whether the restored refresh management information is correct, based on the track shift table 500 stored in the flash ROM 114 during the startup process at power-on. Then check as follows.

  First, the CPU 115 refers to the relative position (relative track position) S (h, c) of the spare track for each track group held in the track shift table 500, so that the relative position S (h, c) of the spare track. Searches for (identifies) a track group that is neither the head position (outermost circumference) nor the final position (innermost circumference) of the track group. The identified track group is a track group that has been the target of the refresh process when the power is shut off before the current power-on.

  Therefore, the CPU 115 determines whether or not the track group number of the specified track group matches the track group number included in the refresh management information restored in the RAM 113. If they match, the CPU 115 determines that the restored refresh management information is correct and the latest refresh management information has not been lost.

  On the other hand, if they do not match, the CPU 115 determines that the restored refresh management information is incorrect and the latest refresh management information is lost. In this case, the CPU 115 determines whether to perform refresh in the forward direction or the reverse direction for the specified track group.

  The refresh direction may be determined in either the forward direction or the reverse direction. In this embodiment, the relative position S (h, c) of the spare track in the specified track group is used to determine this direction. More specifically, the number of tracks Nf and Nb to be refreshed obtained from the relative position S (h, c) of the spare track when the forward refresh and the reverse refresh are assumed, respectively. A magnitude relationship is used. Here, if Nf is Nb or less, forward refresh is determined, and if Nb is less than Nf, reverse refresh is determined.

  The determined refresh direction may be different from the original direction. Therefore, in this embodiment, even when all the refreshes for the identified track group are completed, the number of times of writing (write counter) W (h, c) associated with the track group is a step described later. It is configured not to be cleared at 816. As a result, when the refresh for the identified track group is completed, the refresh process for the track group is continuously executed while switching the refresh direction.

  Note that, for example, a 1-bit direction flag D (h, c) indicating the refresh direction is displayed in the relative position S (h, c) of the spare track for each track group held in the track shift table 500 stored in the flash ROM 114. ) May be added. That is, the track shift table 500 may hold the relative position S (h, c) and the direction flag D (h, c) of the spare track for each track group. If such a configuration is applied, it is possible to specify the direction of the refresh process that has been executed at the time of power-off only by adding information of 1 bit per track group. This eliminates the need for a power shutdown measure that involves determining the refresh direction as described above.

  In step 805, if valid refresh management information is stored in a predetermined area of the RAM 113, the CPU 115 determines the direction of the refresh process based on the state of the direction flag included in the refresh management information. do it. When branching from step 802 to step 805, such a state is obtained.

  However, if valid refresh management information is not stored in a predetermined area of the RAM 113, that is, if the refresh management information is cleared, this means that a new track refresh is performed. It is necessary to check the direction of processing. When the process proceeds from step 802 to step 805 via steps 803 and 804, such a state is obtained. In this case, the relative position S (h, c) of the spare track in the target track group acquired in step 805 is always either the head position (outermost circumference) or the final position (innermost circumference) of the track group. It is in.

  Therefore, if the relative position S (h, c) of the spare track in the target track group is at the innermost circumference of the track group, the CPU 115 sets the direction of the refresh process for the track group as the forward direction. Is determined (step 805). On the other hand, if the relative position S (h, c) of the spare track in the target track group is at the outermost periphery of the track group, the CPU 115 sets the direction of the refresh process for the track group as the reverse direction. (Step 805). In step 805, the CPU 115 stores refresh management information including a direction flag indicating the determined refresh direction in a predetermined area of the RAM 113.

When executing the step 805, the CPU 115 refers to the direction flag included in the refresh management information stored in the RAM 113 as described above to determine whether the direction of the refresh process is the forward direction (step 806)
First, a case where the direction of the refresh process determined in step 806 is the forward direction will be described. In this case, the CPU 115 sets a track adjacent to the spare track on the outer peripheral side of the disk 101 as a refresh target track (target track), and performs data read control for reading data of the adjacent track by the head 102. (Step 810). Next, the CPU 115 executes data write control for writing the read data to the spare track via the HDC 110 (step 811). Thereby, the data of the track (target track) adjacent to the spare track on the outer peripheral side is refreshed on the spare track.

  When the CPU 115 confirms that the writing of the data of the target track to the spare track (step 811) has been completed normally, the spare track in the track group that is currently being refreshed and is held in the track shift table 500 The relative position S (h, c) is shifted to the outer periphery of the disk 101 by one track (step 812). Thereby, the spare track is switched. When the track shift table 500 is updated (switching of spare tracks), the refresh for one track is completed.

  In both the state immediately before and after the update of the track shift table 500 (step 812), the data written in the target track is the track (original track) and the data write destination (the track). On the inner circumference side of the disk 101). However, when the writing of the data of the target track to the spare track (step 811) is normally completed, it is not necessary that the same data exist in both tracks.

  Therefore, in step 812, the relative position S (h, c) of the spare track is changed as described above so that the track treated as the target track immediately before the update of the track shift table 500 is switched to the new spare track. It is shifted to the outer periphery side by one track. In this case, in the next refresh process (steps 810 and 811), the data of the adjacent track on the outer peripheral side on the disk 101 is written in the track that has become a new spare track. If the track shift table 500 (track shift table 500 stored in the flash ROM 114) is updated in a manner that takes into account the power interruption during the update as described above, the spare track can be obtained by the power interruption during the update. There is no failure in switching.

  When the refresh for one track is completed, the CPU 115 determines whether there is a request to interrupt the refresh process (step 813). Reception of a command from the host system 200 is one of the conditions for determining a refresh process interruption request. When the host block 123 in the HDC 110 receives a command from the host system 200, the HDD 100 immediately transitions to the busy state by the hardware function of the host block 123. At the same time, a flag (busy flag) indicating the busy state is set. Therefore, the CPU 115 checks the state of the busy flag in step 813.

  The busy flag is set regardless of whether the refresh operation is being executed. Therefore, the response time of the command recognized from the host system 200 is the sum of the time from when the command is issued until the refresh being executed is completed and the original command execution time. For this reason, when a command is received from the host system 200, if all the refreshes for the target track group are waited for, the response to the command is lowered.

  Therefore, in this embodiment, in order to avoid a decrease in responsiveness to a command, it is determined whether or not there is a refresh processing interruption request as described above every time the refresh for one track is completed (step 813). . If there is a request for interrupting the refresh process (step 813 is YES) as in the case where a command is received from the host system 200 during the execution of refresh for one track, the CPU 115 branches to step 817, The refresh process is immediately interrupted and step 606 (see FIG. 6) is terminated. When CPU 115 completes step 606, it executes step 607. Also in this step 607, the CPU 115 performs the same process as the step 813 in order to quickly branch to the step 603 and start the process for the command from the host system 200.

  On the other hand, if there is no request for interrupting the refresh process (NO in step 813), the CPU 115 determines whether all the refreshes for the target track group have been completed (step 814). This determination is made depending on whether the relative position S (h, c) of the new spare track is at the head position (outermost circumference) in the target track group. The reason is that, when the refresh direction is the forward direction, when the relative position S (h, c) of the spare track is the head position (outermost circumference) in the target track group, the refresh for all the tracks in the group is performed. This is to mean the completion of.

  If the relative position S (h, c) of the new spare track is not at the head position (outermost circumference) in the target track group, and therefore refreshing of all the tracks in the group has not been completed (NO in step 814). ), The CPU 115 branches to step 810 to perform refresh for a new target track, that is, a track adjacent to the new spare track on the outer peripheral side on the disk 101. On the other hand, if the refresh for all the tracks in the target track group has been completed (YES in step 814), the CPU 115 branches to step 815 so that the processing loop (track group processing loop) of steps 810 to 814 is performed. Exit.

  Next, the case where the refresh direction determined in step 806 is the reverse direction will be briefly described. In this case, the CPU 115 performs steps 820 to 824. These steps 820 to 824 are reverse refresh processes corresponding to the forward refresh processes of steps 810 to 814, respectively, and are the same as the forward refresh processes except that the direction is reversed. As a result of the reverse refresh process, the relative position S (h, c) of the new spare track becomes the final position (innermost circumference) in the target track group, and as a result, the completion of the refresh for all the tracks in the group is determined. Then (step 824), the CPU 115 branches to step 815 to exit the processing loop of steps 820 to 824.

  In step 815, the CPU 115 clears the refresh management information stored in the RAM 113 to indicate that the refresh process for the target track group has been completed and no track group is currently being processed. By clearing the refresh management information, it is possible to correctly perform the determination in step 802 and step 806 thereafter.

  Next, the CPU 115 initializes (that is, clears) the number of writes W (h, c) held in the write count table 400 in association with the track group for which the refresh process has been completed (step 816). The write count W (h, c) for each track group held in the write count table 400 indicates the number of times of execution of the write operation to the track group. The number of write times W (h, c) correlates with the degree of data deterioration in the track group associated with the number of write times W (h, c). Therefore, in this embodiment, the number of write times W (h, c) is treated as the degree of data deterioration for convenience. Immediately after the refresh process for the track group is completed, there is no data deterioration in the group. In order to reflect this in the write count W (h, c) associated with the track group for which the refresh process has been completed, the write count W (h, c) is initialized to 0 as described above. .

  When executing step 816, the CPU 115 again branches to step 802 and starts processing for the next track group. In this case, the branch destination from step 802 is always step 803, and processing including an operation for searching for a track group that requires refresh is performed in the same manner as described above. If there is no request for interrupting the refresh process, the refresh process for the track group (track refresh process) is terminated only when there is no track group that requires the track group (NO in step 804).

[First Modification]
Next, a first modification of the above embodiment will be described. A feature of the first modification is that the track shift table 500 is stored in the RAM 113 in the same manner as the write count table 400. Such a configuration is essential, for example, when the flash ROM 114 shown in FIG. 1 cannot frequently rewrite data, or when a ROM that cannot be rewritten is used instead of the flash ROM 114.

  Also in the first modification, the track refresh process according to the flowchart of FIG. 8 is executed. However, in the first modification, for example, immediately after step 805, the refresh management information and the track shift table 500 stored in the RAM 113 are stored in a predetermined area of the disk 101 under the control of the CPU 115. At this time, the in-execution flag included in the refresh management information is set, indicating that the refresh process for the track group specified by the track group number included in the management information is being executed. . In the first modification, for example, immediately after step 816, the execution flag included in the refresh management information stored in the area of the disk 101 is reset under the control of the CPU 115. Thereafter, the track shift table 500 stored in the RAM 113 at that time is stored in the area of the disk 101.

  According to the first modified example, even if the power supply is unexpectedly shut down during the refresh process, an in-execution flag indicating that the execution is in progress remains in a predetermined area of the disk 101 when the HDD 100 is started next time. You can detect it by checking if it is.

  In general, ECC (Error Correcting Code) data is added by the HDC 110 to data written in each sector arranged in each track of the disk 101. In the first modification, a value generated based on the virtual cylinder number of the track in which the sector to which the ECC data is added is used is used as the seed value of the ECC data. More specifically, as a seed value of ECC data, a value generated based on the virtual cylinder number and head number of the track in which the sector to which the ECC data is added is arranged, and the sector number of the sector. For example, a value obtained by concatenating a virtual cylinder number, a head number, and a sector number is used.

  As described above, in the configuration in which the virtual cylinder number is used for generating the seed value of the ECC data, even if the relative position of the spare sector held in the track shift table 500 is incorrect, the HDC 110 can detect the error. . The reason is as follows. First, it is assumed that the data of the track adjacent to the erroneous spare sector is read in step 810 or 820. In this case, since the seed values are different, an ECC error (seed error) occurs in the HDC 110. Thereby, it can be detected that the relative position of the spare sector is incorrect.

  Now, it is assumed that an in-execution flag indicating that an execution is in progress remains in a predetermined area of the disk 101 when the HDD 100 is activated. In this case, the CPU 115 determines (detects) that an unexpected power shutdown has occurred during the execution of the refresh process. The refresh management information including the in-execution flag and the track shift table 500 stored in the predetermined area of the disk 101 and the track shift table 500 are read from the predetermined area of the disk 101 and stored (restored) in the predetermined area of the RAM 113. )

  Here, the relative position of the spare track held in the track shift table 500 in association with the track group for which the refresh process has been executed when the unexpected power shutdown occurs is incorrect due to the power shutdown. there is a possibility. Therefore, in the first modification, when the relative position of the spare track is incorrect, it is corrected to the correct relative position as described below.

  First, from the track shift table 500 stored in a predetermined area of the RAM 113, the relative position of the spare track associated with the track group for which the refresh process has been executed when the power interruption occurs is acquired. This track group is indicated by a track group number included in the refresh management information stored in a predetermined area of the RAM 113.

  Next, a data read operation is performed assuming that the virtual cylinder number is the same as the actual cylinder number for each track in the target track group. By this data read operation, the boundary between the track having the matching seed value and the track not having the seed value (track having a seed error) is specified. Of the two tracks adjacent to this boundary, the track in which the seed error has occurred is determined as the correct spare track. If the determined relative position of the spare track is different from the relative position of the spare track acquired from the track shift table 500, the acquired relative position of the spare track is incorrect. In this case, the CPU 115 corrects the value of the relative position of the erroneous spare track held in the track shift table 500 to the value of the determined relative position of the spare track.

  Here, binary search is used to determine the order of reading data from each track in the target track group in order to speed up the search. The sector to be read in the track is not particularly limited, and may be a sector at a predetermined relative position in the track, for example. In addition, the sector with the least waiting for rotation in the track may be dynamically selected.

  According to the first modified example, even if an unexpected power interruption occurs, this can be detected at the next startup of the HDD 100. In addition, according to the first modification, the relative position information of the spare track held in the track shift table 500 can be restored without storing the track shift table 500 in the flash ROM. Therefore, according to the first modified example, even if the flash ROM 114 cannot be used to store the track shift table 500 or the HDD does not have the flash ROM 114, the refresh process based on the track shift table 500 is performed. Can be realized.

[Second Modification]
Next, a second modification of the above embodiment will be described. A feature of the second modified example is that the refresh unit (that is, the unit for reading / writing data for refresh) is optimized in order to reduce the rotation waiting time in the reverse refresh process. is there.

FIG. 9 is a conceptual diagram of optimization of refresh units for the purpose of shortening the rotation waiting time by reverse refresh processing.
As shown in FIG. 9, a track skew is applied between adjacent tracks arranged on the disk 101 to adjust the position of the head sector in the track. The value of the track skew is determined so as to minimize the rotation waiting time in the forward access, that is, the sequential access in the direction in which the LBA increases. That is, after accessing the last sector of a track on the disk 101, the time required to move the head 102 to the next (adjacent) track is the minimum track skew.

  Therefore, when the track refresh is performed in the forward direction, the track skew is optimized, and the rotation waiting time is small. However, when the track refresh is performed in the reverse direction, the rotation waiting time increases remarkably.

  FIG. 9A shows a trajectory that the head 102 draws on the disk 101 in the reverse track refresh (that is, the reverse refresh process performed when the refresh unit is one track as in the above embodiment). . As shown in FIG. 9A, in the track refresh in the reverse direction, the head 102 sequentially draws a locus 903, a locus 902, and a locus 901 on the disk 101.

  A locus 903 indicates a locus drawn by the head 102 in a read operation (read access) for reading data from the target track, and a locus 901 indicates a head in a write operation (write access) for writing the read data to the spare track. The locus | trajectory which 102 draws is shown. Here, a long rotation waiting time corresponding to the length of the locus 902 occurs.

  In the second modification, the refresh unit (refresh amount per refresh process) is optimized in order to shorten the rotation waiting time as indicated by the locus 902 in FIG. 9A. FIG. 9B shows, as an example, a trajectory that the head 102 draws on the disk 101 in the reverse refresh process performed when the refresh unit is optimized to exceed 2 tracks and below 2 tracks. . In the example of FIG. 9B, the head 102 sequentially draws a locus 913, a locus 912, and a locus 911 on the disk 101.

  A locus 913 indicates a locus drawn by the head 102 in the read access for reading data from the target track, and a locus 911 indicates a locus drawn by the head 102 in the write access for writing the read data to the spare track. The rotation waiting time corresponding to the length of the locus 912 is shortened compared to the rotation waiting time corresponding to the length of the locus 902 in the example of FIG.

  Depending on the value of the track skew, the refresh unit may be less than one track, and such optimization is possible. FIG. 9C shows a trajectory that the head 102 draws on the disk 101 in the refresh process in the reverse direction performed when the refresh unit is optimized to be less than one track. In the example of FIG. 9C, the head 102 sequentially draws a locus 923, a locus 922, and a locus 921 on the disk 101.

  A locus 923 indicates a locus drawn by the head 102 in the read access for reading data from the target track, and a locus 921 indicates a locus drawn by the head 102 in the write access for writing the read data to the spare track. The rotation waiting time corresponding to the length of the locus 922 is shortened compared to the rotation waiting time corresponding to the length of the locus 902 in the example of FIG. Further, when the refresh unit is increased, it is necessary to secure a large buffer area in the buffer RAM 111 according to the size. Therefore, the refresh unit may be reduced for the purpose of reducing the buffer area.

Next, a specific method for calculating the refresh unit will be described.
As described above, FIG. 9B shows a trajectory that the head 102 draws on the disk 101 in the reverse refresh process performed when the refresh unit is optimized to exceed one track and less than two tracks. . One track and two tracks correspond to one rotation and two rotations of the disk 101, respectively. Therefore, the refresh unit exceeding 1 track and 2 tracks or less can be rephrased as the refresh unit exceeding 1 rotation of the disk 101 and 2 rotations or less.

Here, the number of revolutions per unit time (one second) of the disk 101 is f [rotations / s], and the time required for the movement (seek operation) of the disk 101 over x tracks is t _x [s]. Further, the access amount (necessary for drawing the locus 913) corresponding to the locus 913, that is, the refresh unit, is represented by the number of rotations u [rotation] of the disk 101. However, u is 1 <u ≦ 2 from the precondition of the refresh unit.

Next, the rotation speed of the disk 101 corresponding to the locus 912 (rotation waiting time) is set to w [rotation]. t _x includes not only the time required for the physical movement of the head 102 but also the setup time for the read / write access at the movement destination of the head 102. The track skew is equal to the time t ₁ [s] required for seeking one track (x = 1). The rotational speed of the disk 101 at this time t ₁ [s] is represented by ft ₁ . That is, the track skew corresponds to the rotation speed ft ₁ .

At this time, from FIG. 9B, the number of revolutions obtained by subtracting the number of revolutions ft ₁ corresponding to the track skew from the two revolutions of the disk 101 for the read access and the rotation waiting time corresponding to the trajectories 913 and 912, respectively. It turns out that it requires. Therefore, the following equation: u + w + ft ₁ = 2 (1)
Is established.

Further, it can be seen from FIG. 9B that w requires at least the number of revolutions required for the movement of the head 102 over two tracks. However, unlike the example of FIG. 9B, the track crossing may occur twice depending on the start position of the track 913, for example, in the case where the start position of the track 913 is on the end side of the head track. There is also a possibility that the trajectory 913 extends over three tracks. For this reason, w is expressed by the following equation: w ≧ ft ₃ (2)
Subject to the restrictions shown in If w does not satisfy this restriction, a waiting for rotation occurs for one extra rotation.

From the above formulas (1) and (2) and the precondition 1 <u ≦ 2 of the refresh unit u [rotation], the refresh unit u [rotation] is expressed by the following formula 1 <u ≦ 2-f (t ₁ + t ₃ ) [Rotation] (3)
Can be obtained as follows. Here, the case where the equal sign holds (that is, when u = 2−f (t ₁ + t ₃ )) is the most efficient, and refreshing is performed within a range not exceeding this (2−f (t ₁ + t ₃ )). Decide on the unit.

In the example of FIG. 9B, when the write access corresponding to the locus 911 is completed, the refresh process for the next refresh unit is started and the read access is performed. The read access start position is a position where the read access corresponding to the locus 913 is completed. The access amount for write access corresponding to the locus 911 is equal to that for read access corresponding to the locus 913. Therefore, the rotational speed of the disk 101 required for the seek operation for moving the head 102 from the position where the write access corresponding to the locus 911 is completed to the start position of the next read access is the rotational speed ft ₁ corresponding to the track skew. equal. Therefore, it is possible to shift to the next refresh unit with the minimum rotation waiting time.

Next, FIG. 9C shows that the head 102 draws on the disk 101 in the reverse refresh process performed when the refresh unit is optimized to be less than one track (less than one rotation) as described above. Represents a trajectory. In this case, as in the case where the refresh unit exceeds one track (one rotation), the refresh unit u [rotation] is expressed by the following equation: 0 <u ≦ 1-f (t ₁ + t ₂ ) [rotation] (4 )
Can be obtained as follows.

In order to more generally express the optimization of the refresh unit, it is assumed that k is an integer equal to or greater than 1, and the refresh unit is optimized to exceed (k−1) rotations and less than k rotations. In this case, the refresh unit u [rotation] is expressed by the following equation: k−1 <u ≦ k−f (t ₁ + t _{k + 1} ) [rotation] (5)
Can be obtained as follows. That is, the unit u of refresh is determined based on the value obtained by adding the t _{k + 1} to _{t 1 (t 1 + t k} + 1). Here, t ₁ corresponds to the value of the track skew as described above, and t _{k + 1} is the head from the position where the read access is completed to the write access start position for writing the data read by the read access. This corresponds to the time required to move 102.

For example, if the HDD 100 has f = 5400 [rpm] = 90 [rotations / s], t ₁ = 2.0 [ms], t ₂ = 2.2 [ms], and t ₃ = 2.4 [ms]. , K is 1,
0 <u ≦ 1-90 × (2.0 + 2.2) × 10 ⁻³ = 0.622 [rotation]
It becomes. When k is 2,
1 <u ≦ 2-90 × (2.0 + 2.4) × 10 ⁻³ = 1.604 [rotation]
It becomes.

  In any case, the access amount in one refresh may be the number of sectors (blocks) that can be accessed without exceeding the number of rotations. In calculating the number of sectors, it is necessary to consider that there is no accessible sector in the track crossing portion included in the access range.

  In addition, this invention is not limited to the said embodiment or its modification example as it is, A component can be deform | transformed and embodied in the range which does not deviate from the summary in an implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment or its modification. For example, you may delete a some component from all the components shown by embodiment or its modification.

  DESCRIPTION OF SYMBOLS 100 ... Magnetic disk apparatus (HDD), 101 ... Disk, 102 ... Head, 113 ... RAM, 114 ... Flash ROM, 115 ... CPU (controller), 201 ... Track, 400 ... Write count table, 500 ... Track shift table.

Claims (3)

In a magnetic disk device that performs data refresh,
A magnetic disk Lud over data track group comprising a preliminary data tracks and a plurality of data tracks that are used in the data refreshing is plurality,
When refreshing the plurality of data tracks in any data track group of the plurality of data track groups, the data recorded in the data track adjacent to the front Ki予 Bei data track in the data track group before being moved to Ki予 Bei data tracks, the adjacent data tracks are repeated switching which is a next preliminary data tracks in response to this, the control to the preliminary data tracks are moved in sequence within the data track group magnetic disk device you and a controller. And a controller for controlling the refreshing of data Lud over data track group comprising a preliminary data tracks and a plurality of data tracks that are used in data refresh is recorded in the plurality is a magnetic disk,
When refreshing the plurality of data tracks in any data track group of the plurality of data track groups, the data recorded in the data track adjacent to the front Ki予 Bei data track in the data track group before being moved to Ki予 Bei data tracks, the adjacent data tracks are repeated switching which is a next preliminary data tracks in response to this, the control to the preliminary data tracks are moved in sequence within the data track group Turkey controller to include a control means for. A data refresh method in the magnetic disk apparatus having a magnetic disk Lud over data track group comprising a preliminary data tracks and a plurality of data tracks that are used in the data refreshing is plurality,
When refreshing the plurality of data tracks in any data track group of the plurality of data track groups, the data recorded in the data track adjacent to the front Ki予 Bei data track in the data track group Go before Ki予 Bei data tracks, the adjacent data tracks is repeatedly switched to the next preliminary data tracks in response to this, the mobile to Lud over data refresh in the preliminary data tracks sequentially in the data track group Method.

Images