[ale] One possible explanation - Re: 9.10 smart errors

[David Ritchie]

Ok, here is how it works.

All modern drives have a number of blocks (typically located on one or
more cylinders of the drive) that contain replacement space for
sectors that are known to have defects. All drives in use today use
error correcting codes to provide correction of a variety of single
and multi-bit errors, along with using retry to make another stab at a
correct
read (but at the cost of a full rotational disk latency). Tools such
as Spinrite can be used to repeate read attempts to attempt to recover
data on drives with sectors with marginal sectors, and while slow, can
often recover corrected data.

When you format a drive on a SCSI drive, you can specify bad sectors
for sparing. These are placed into the G-list, which is a list of bad
sectors that have been discovered since the original factory format.
This is known as the grown defect list. There is also a permanent
defect list called the P-list, that lists all permanent defects
discovered during the manufacturing process for the drive. You can
delete the G-list on a drive, but not the P-list. Normally, unless you
go out of your way to turn it over via mode commands, sectors with
excess retry, failed ECC etc. are spared automatically and added to
the G-list after attempting to move the data to a new track.

The only reason I would expect to see a modern drive with real bad
sectors is that we have exhausted the spare pool
and it is not possible to reallocate space. The drive is probably
toast in that case anyway.

-- Dave Ritchie