Re: chkdsk /r and surface scan

On Sat, 18 Jun 2005 18:18:34 -0700, Anthony Ewell <aewell@xxxxxxxx>

>Thank you for the excellent answer. I have used the "/r"
>option frequently but have never seen any errors on the
>last pass (open space). The drives worked a lot better
>afterwards. This explains it.

> It is also nice to know that (S)ATA's are automatically
>swapping out bad sectors.

Not really, no.

If your HD's dying, do you want to know that early, when the first bad
sector comes to light, or when it dies and eats your data?

The fact that the drive "worked a lot better afterwards" suggests that
the process did indeed "fix" bad sectors. That there were bad sectors
to be found, is reason to immediately evacuate your data and get off
that HD. If ChkDsk "fixed" (i.e. covered up) bad sectors and did not
tell you it did so, then that's reason enough to not trust ChkDsk.

Think about it - if you ever see a disk error, it means:
1) The HD's firmware defect management couldn't cover it up
2) The NTFS drivers couldn't cover it up

Both (1) and (2) work in the same way, but at different levels of
awareness. (1) works within the HD itself, and is invisible to the
OS; your only window into the process is S.M.A.R.T. records, that only
a detailled S.M.A.R.T. reporter will show you (many HD vendor's tools
will just show you a glib "S.M.A.R.T. says the HD's OK").

In contrast, (2) is - or should be - visible to the OS; you'd expect
bad clusters to be marked out of use within the file system, and you'd
hope that ChkDsk (or AutoChk) would log the details of this; not just
new ones, but count of existing ones. However, you may find there's
no way to get the OS to show you this info.

Now it may be impossible to switch off the HD's own firmware defect
management. If so, then whenever NTFS's driver code, or a surface
scan via ChkDsk /r, "sticks" on a failing sector, the firmware itself
is probably trying to hide it. Chances are these two competing
processes will tangle up, causing a long delay that looks like a hang.

Think about it; let's say NTFS tries to read the sector 100 times, and
the firmware tries 100 times, too. Each time NTFS tries, this causes
the firmware to try 100 times, making it 10 000 times in all, but
here's the thing; a successful read is 100 times more likely to happen
while the firmware's retrying, than when the NTFS is retrying, unless
the sector is so dead that all the retries in the world can't help.

Both processes try to do the same thing; read the data out of the
dying sector, verify the data's OK via an integrity check (CRC or
something), and if it's OK, write this data somewhere else that
doesn't suck. But a bit pun might pass CRC with bad data, and what
happens if the data can't be read at all? Just copy nothing to the
new place, swap the pointers, and hope no-one notices?

If the firmware does the remapping, the OS is unaware of it, unless it
gets suspicious about the number of retries or has some sort of
real-time awareness. If it does, it may remap the data itself anyway,
but when the OS does it, it has to amend the volume's file system
structure, and somehow mask the bad sector from re-use.

What this means is you could have a really sick HD that looks "OK" to
ChkDsk or Scandisk, because all defects are hidden by the firmware.
There may be a substantial number of bad sectors remapped at the HD's
firmware level, and not a single adverse finding by ChkDsk or
Scandisk, but because better sectors are used, the HD speeds up.

To me, that's a clear call to head for the lifeboats.

You may want to know what I say that, if these bad sectors are
"fixed". Well, that's because while every HD may have initial
manufacturing defects that are mapped out, *new* defects are
significant. Why would there be new defects? Surface wear? If so,
the internal airspace is probably polluted with abrasive debris; where
the heads go, further wear is likely to follow. Or is it something
awry with the head quality or positioning? Same mileage; if it's gone
from fine to bad, it will progress from bad to worse.

You need to know that, but everything conspires to hide that from you,
so you can stagger on until the HD goes out of warranty.



>------------------------ ---- --- -- - - - -
Forget http://cquirke.blogspot.com and check out a
better one at http://topicdrift.blogspot.com instead!
>------------------------ ---- --- -- - - - -
.