Seagate ST336918N, ST318418N, ST318438LW manual Defect and error management, Drive internal defects

Barracuda 36ES2 Product Manual, Rev. B 33

7.0Defect and error management

The drive, as delivered, complies with this specification. The read error rate and specified storage capacity are not dependent upon use of defect management routines by the host (initiator).

Defect and error management in the SCSI system involves the drive internal defect/error management and SCSI systems error considerations (errors in communications between Initiator and the drive). Tools for use in designing a defect/error management plan are briefly outlined in this section, with references to other sections where further details are given.

7.1Drive internal defects

During the initial drive format operation at the factory, media defects are identified, tagged as being unusable, and their locations recorded on the drive primary defects list (referred to as the “P” list and also as the ETF defect list). At factory format time, these known defects are also reallocated, that is, reassigned to a new place on the medium and the location listed in the defects reallocation table. The “P” list is not altered after factory formatting. Locations of defects found and reallocated during error recovery procedures after drive shipment are listed in the “G” list (defects growth list). The “P” and “G” lists may be referenced by the initiator using the Read Defect Data command (see the SCSI Interface Product Manual, part number 75789509).

7.2Drive error recovery procedures

Whenever an error occurs during drive operation, the drive, if programmed to do so, performs error recovery procedures to attempt to recover the data. The error recovery procedures used depend on the options previ- ously set up in the error recovery parameters mode page. Error recovery and defect management may involve the use of several SCSI commands, the details of which are described in the SCSI Interface Product Manual. The drive implements selectable error recovery time limits such as are required in video applications. For addi- tional information on this, refer to the Error Recovery Page table in the SCSI Interface Product Manual, which describes the Mode Select/Mode Sense Error Recovery parameters.

The error recovery scheme supported by the drive provides a means to control the total error recovery time for the entire command in addition to controlling the recovery level for a single LBA. The total amount of time spent in error recovery for a command can be limited via the Recovery Time Limit bytes in the Error Recovery Mode Page. The total amount of time spent in error recovery for a single LBA can be limited via the Read Retry Count or Write Retry Count bytes in the Error Recovery Mode Page.

The drive firmware error recovery algorithms consist of 11 levels for read recoveries and 15 levels for writes.

Table 3 equates the Read and Write Retry Count with the maximum possible recovery time for read and write recovery of individual LBAs. The times given do not include time taken to perform reallocations, if reallocations are performed. Reallocations are performed when the ARRE bit (for reads) or AWRE bit (for writes) is one, the RC bit is zero, and the Recovery Time Limit for the command has not yet been met. Time needed to perform reallocation is not counted against the Recovery Time Limit.

The Read Continuous (RC) bit, when set to one, requests the disc drive to transfer the requested data length without adding delays (for retries or ECC correction) that may be required to insure data integrity. The disc drive may send erroneous data in order to maintain the continuous flow of data. The RC bit should only be used when data integrity is not a concern and speed is of utmost importance. If the Recovery Time Limit or retry count is reached during error recovery, the state of the RC bit is examined. If the RC bit is set, the drive will transfer the unrecovered data with no error indication and continue to execute the remaining command. If the RC bit is not set, the drive will stop data transfer with the last good LBA, and report a “Check Condition, Unrecovered Read Error.”