hard disk integrity check

[Alan McKinnon]

On Monday 17 April 2006 00:22, Yorvik wrote:
> Gary W. Swearingen wrote:
> > I'll propagate a rumor I heard: Use of "badblocks" has been
> > obsolete since drives started to remap their own bad blocks many
> > years ago. You won't find bad blocks until after your drive has
> > found so many that it's better used as a doorstop.
> >
> > I beg to be corrected by someone with real knowledge.
>
> I was told a couple of years back, that if you can find bad blocks
> with 'normal user software' the drive has had it and may as well be
> chucked.

That's not true. There's many urban myths surrounding disks, and this 
is one of them.

> Personally, I can remember when harddisks had labels on them
> listing the bad blocks.

Those were MFM drives. Drive manufacturers soon got fed up with the 
support calls from users as to why they shipped faulty disks - ALL 
drives have errors. Solution - disguise the errors. Modern drives 
keep a small percentage of space free for bad blocks and as the 
firmware picks up failing sectors, it remaps their location into this 
unused space. The whole cyclinder/sector scheme is a pure abstraction 
anyway, so this works well.

When you run badblocks and pick up a few errors, all you are doing is 
beating the drive firmware to the same job. When you find it with 
user software, it means that the firmware has given up trying. 
Reading and writing to a disk is a tricky job (almost but not quite 
entirely unlike reading and writing to RAM), and the firmware is 
limited in what it can get the heads to do. The end result is that it 
gives up easily, which opens up a nice market for third party 
software. This is most of what disk data recovery is about. Spinrite 
is a good example and http://www.grc.com explains it all nicely as 
long as you can read past Steve Gibson's hyped opinions.

What you should be worried about is the *rate* at which bad sectors 
are being produced. Once this passes a certain well-defined amount, 
then the drive is becoming statistically more likely to fail. This is 
what SMART is all about.

So while what you were told is not completely wrong, it's over 
simplified by a very big margin

-- 
If only you and dead people understand hex, 
how many people understand hex?

Alan McKinnon
alan at linuxholdings dot co dot za
+27 82, double three seven, one nine three five