> Yes, that's how I've seen it work. Take a drive with one bad sector,
> overwrite that sector, and the pending count drops to zero.
Agreed: this is what I have also seen.
> > In the drive in question, the pending sector count was large, the
> > reallocated sector count was zero, and the drive HAD shown one read
> > scan
> > error, but this error had disappeared in the most recent read scan,
> > which
> > showed zero read errors.
> > Since the most recent read scan showed no errors, this would imply to
> > me
> > that any bad sectors had either been reallocated or fixed by rewriting
> > the
> > track. So in this case the pending sector count ought to be zero.
> > What's wrong with this logic?
> My best guess: The sector is read as invalid, and so it gets put in the
> pending list, but the read scan later reads the same sector and sees it
> as OK (but, I assume, doesn't remove it from the pending list). When
> the sector is written, the drive successfully rewrites the sector, so
> it gets removed from the pending list and not reallocated.
> That matches the original description, which is that the trouble
> happens at startup but not later.
Your explanation is consistent with what's observed.
I had assumed that the sector would be removed from the pending sector
list as soon as the drive can read that sector (ie, as part of the read
scan). Your explanation matches the observations better. To restate it,
a sector will ONLY be removed from the pending sector list on a WRITE,
never on a READ. The algorithm for removing it on a WRITE is that first,
you try to write to the (supposedly) bad sector, and then immediately do a
verifying READ. If that suceeds, you remove the sector from the pending
sector list without reallocating it. If the verifying READ fails, then
you reallocate the sector to a spare sector and remove it from the pending
Geoff, do you know if all vendors do this? I think I've seen this behavior
consistently on IBM and IBM/Hitachi drives. I think that Maxtor's might
remove the sector from the pending list as soon as it can be read (as part
of an offline scan) but I'm not sure. Geoff, if you don't know I'll ask a
contact at Maxtor about this.