Scanning a disk for defects using badtrk

Use badtrk(ADM) to scan IDE, EIDE, UDMA, ESDI, and SCSI disks for defective tracks. It maps any flawed tracks to good ones elsewhere on the disk. It also creates a table of all the bad tracks on your hard disk.

When you first install a SCSI disk, badtrk creates a table of bad blocks in the partition. On SCSI disks, badtrk tries to use spare disk blocks that are maintained by the disk controller as replacements for bad blocks. If bad blocks cannot be mapped out in this way, the disk driver maps out bad blocks using the spare blocks and the bad block table in the disk partition. You can force it to use this table by specifying the -O option to badtrk. You can also enable Automatic Read/Write Remapping (ARR/AWR) for the entire SCSI disk if the disk controller supports this feature. Any defects that arise will be remapped without notifying you.

badtrk can:

• display the current bad track/block table. This is an example of a bad track table with a single entry:

     Defective Tracks
     
  
     +-----------------------------------------------+
     |     Cylinder   Head     Sector Number(s)      |
     +-----------------------------------------------+
     |1.     190       3         12971-12987         |
     +-----------------------------------------------+
  

• scan the disk for flaws. The scan can be ``quick'' or ``thorough'', destructive or non-destructive. Typical scan rates are 18MB per minute for quick, and 6MB per minute for thorough. Do not choose a destructive scan if you want to preserve existing data on the disk.

  Whenever badtrk finds a defective track, it displays its location on the disk, for example:

  WARNING : wd: on fixed disk ctlr=0 dev=0/47 block=31434 cmd=00000020
      status=00005180, sector = 62899, cylinder/head = 483/4
  

  You can interrupt the scan at any time and return to the main badtrk menu.

• add entries to the current bad track/block table.

  On some IDE, EIDE, and UDMA drives that remap the disk geometry, the cylinder, head, sector reported for a bad sector or block may not correspond to the actual physical values of these quantities. In such a case, you should determine an absolute block address for the sector(s) or block(s) to be remapped using the formula:
  
  logical address = ((cylinder+1) x (head+1) x (sectors per track)) + sector

  Note that this assumes that the first cylinder, head, and sector are all numbered from 0.

• delete individual entries from the current bad track/block table.

• delete all entries from the bad track/block table.

• enable Automatic Read/Write Remapping on SCSI disks that support this feature. Any bad blocks that develop on the disk will be reallocated automatically to the defect list managed by the disk controller.

When installing a new disk, you should perform a thorough destructive scan on the complete UNIX system partition. It may take several hours to scan a large hard disk.

See hd(HW) for a description of hard disk naming conventions.

When using mkdev hd to install a hard disk, if badtrk finds a flaw in the first few tracks of the UNIX system partition, it returns you to fdisk. You can then repartition the disk to exclude the defective tracks from any partition. When you leave fdisk, badtrk runs again to allow you to scan the disk for further flaws. This process continues until badtrk finds no flaws in the first few tracks. You may have to experiment to determine how many tracks to exclude.

When you quit badtrk while first installing a disk, it prompts you for the number of tracks to reserve as replacements for flawed ones. Allocate at least as many tracks as badtrk recommends. This number is based on the current number of bad tracks plus an allowance for tracks that may go bad. If you ever exceed the number of allocated bad tracks, you must reinstall the hard disk.