notes on bad sector recovery

[Eric Gisin]

I have a Maxtor 20GB drive that developed bad sectors some time ago due to a
power glitch. It contains a FAT32 volume, which does not mount in Win2K because
of bads in the FAT. Findpart ran for an hour before I gave up on it.

My first idea was to copy the FATs to a file and work on that. I used GNU dd:
dd if=//./R: of=image conv=noerror,sync count=40000
That took 3 hours. I expect bad sectors to be replaced with zeros or garbage,
but instead I got the data from the bad sectors!

Do most IDE drives return the bad data with single sector Read DMA? I know Read
Long will (if supported), but you need do this from DOS with your own driver.

There were 1807 bads in the 2*19521 FATs. No new bads, but SMART reports
reallocations.

I will followup on how I processed the image.

 

[Folkert Rienstra]

I have a Maxtor 20GB drive that developed bad sectors some time ago due to a
power glitch.

All 1807 of them? No way.

It contains a FAT32 volume, which does not mount in Win2K because
of bads in the FAT. Findpart ran for an hour before I gave up on it.

My first idea was to copy the FATs to a file and work on that.

One of Svend's utes can do that for you.

I used GNU dd:
dd if=//./R: of=image conv=noerror,sync count=40000
That took 3 hours. I expect bad sectors to be replaced with zeros or garbage,
but instead I got the data from the bad sectors!

Do most IDE drives return the bad data with single sector Read DMA?

Spec says "The amount of data transferred is indeterminate" so maybe some
manufacturers may interpret that to mean that bad data can be delivered.

This was discussed also october 2002 in Re: Sector Read errors...

I know Read Long will (if supported), but you need do this from DOS with
your own driver.

Well, unless you specifically know what app uses which instruction you probably
have to write your own app too. Apparently Read Sectors and Read Multiple
will also transfer a bad block.

 

[Arno Wagner]

I have a Maxtor 20GB drive that developed bad sectors some time ago due to a
power glitch. It contains a FAT32 volume, which does not mount in Win2K because
of bads in the FAT. Findpart ran for an hour before I gave up on it.

My first idea was to copy the FATs to a file and work on that. I used GNU dd:
dd if=//./R: of=image conv=noerror,sync count=40000
That took 3 hours. I expect bad sectors to be replaced with zeros or garbage,
but instead I got the data from the bad sectors!

You should not use dd for that. It has no error handling except abort.
However I suspect that you don't have defect secors (or the disk
remapped them), since dd should abort on any error. Maybe this is
an artefact from using a Unix-tool on Windows, but I doubt it.

If you want to copy a disk with true bad sectors, use dd_rescue
and Linux, it gives you the option to write nothing or zeros for
unreadable source sectors and other than dd it will not abort on read
errors.

Do most IDE drives return the bad data with single sector Read DMA?
I know Read Long will (if supported), but you need do this from DOS
with your own driver.

Most IDE drives return an error when a defect sector is read.

There were 1807 bads in the 2*19521 FATs. No new bads, but SMART reports
reallocations.

Are you talking about Filesystem-marked bad sectors? These don't stay
bad on a modern disk. The disk will remap them as soon as they
are read correctly once or are written once.

Arno

 

[Folkert Rienstra]

You should not use dd for that.

It has no error handling except abort.

Did you even consider for a split second that maybe that is exactly what is wanted?

However I suspect that you don't have defect secors (or the disk
remapped them), since dd should abort on any error.

And here we just hear you say "It has no error handling except abort".
Guess you just lied then.

Maybe this is an artefact from using a
Unix-tool on Windows, but I doubt it.

If you want to copy a disk with true bad sectors, use dd_rescue
and Linux, it gives you the option to write nothing or zeros for
unreadable source sectors

And throw away your only source for comparing what data
bytes in the Fats are save and not need reconstruction?

and other than dd it will not abort on read errors.


Most IDE drives return an error when a defect sector is read.

They all should. If they don't you should ask your money back.

Are you talking about Filesystem-marked bad sectors?

Rant, rant, rant .... Think Arnie, would Findpart run for hours if the
bad sectors were only recorded in the Fat. Why would he want to
compare Fats if the Fats were OK. Think, Arnie, think.
I know that it is mighty difficult for you, but for once, try.

And before you want to try explain how the bad sectors in the
FAT may have been recorded in the FAT ....... Think Arnie, think.

These don't stay bad on a modern disk.

Rotflol. Arnie tries to teach Eric a lesson. Go Arnie, go.

The disk will remap them as soon as they
are read correctly once or are written once.

Gee Arnie, guess who told you that.
And no, the drive may just rewrite the sector without replacement.
And no, a bad sector that hasn't been marked as such isn't replaced either.