Prev: Q about minimized documents and Dock in Snow Leopard
Next: Mac - Windows filesharing search issue
From: MartinC on 24 Jul 2010 06:43
David Empson wrote:

> You could try using the erase with zero data option in Disk Utility, but
> it might fail if it encounters a bad sector.

To my best of knowledge, quite the opposite - if a bad sector gets detected
while *writing* the block, the drive firmware will re-allocate it.

Sadly, the firmware will not do that while reading, so if a single block
dies, every attempt to read it will just throw errors, only if you try to
write data to it will re-allocate it.

Performing a "low level" format writing zeroes should dispose of all damaged
blocks. There are two possibilities:

1) If the drive is mechanically fine, but only some small spots on the
surface damaged, then the drive will be fine afterwards. A single damaged
block does not automatically mean that the whole drive is dying.

2) If the drive is mechanically defective and actually *caused* the block
damage, then zero-ing the whole drive will probably rip it in pieces.

One way or the the other, you should know afterwards.

From: MartinC on 25 Jul 2010 04:04
Nick Naym wrote:

> You guys are losing -- nah: have already lost! -- me!
>
> But your "re-allocate" comment sounded familiar: As I indicated in my
> initial post about this problem,
>
>
> � ...I decided to reformat/reinitialize the drive. However, my repeated
> attempts to do so fail: After about 10 seconds of churning after I click the
> �partition� button, DU gives up and issues the error message: �Partition
> failed with the error: Could not unmount disk.�
>
> Since the drive has a USB interface as well as FW, I also tried
> reformatting/reinitializing via a USB port. That too was unsuccessful, but
> the error message was different: �Partition failed with the error: POSIX
> reports: The operation couldn�t be completed. Cannot allocate memory.� �
>
>
> So, apparently re-allocation was an issue via USB.
>
>
> (Any effort on your part to explain what the hell is going on without losing
> me would be greatly appreciated.)

Sorry for the confusion... I try to explain (describing it a bit simpler as
it actually is... ;-)

The harddisk has zillions of blocks that are actually terribly sensitive -
there is a real chance that some few might blow one day, even if the drive
remains OK otherwise. In order to deal with that problem, the drive "itself"
(i.e. the firmware) has a strategy...

It keeps a lot of hidden blocks that are not used at all - you will not see
them and you can't access them.

Every time when you either read or write a block, the drive itself will know
if it physically worked. Now the magic trick: If the drive *writes* a block
and it physically fails because that single block got damaged since last
time, it will silently swap the block with with one of those hidden
"replacement" blocks - so the bad one will be marked out and never used
again and the new one will be the substitute. This is called
"re-allocation".

You, as the user, will probably never even realize that this just happens -
you wrote a file, it was written, fine!

However, the drive will NOT do this when you *read* a block and the reading
fails. Why? Because if it can't read the block, there is no way to get a
replacement block with the same content.

So if you have an ordinary file where one of the blocks got damaged, if you
try to open it with an application or copy it in the Finder you get a box
"Sorry this file can't be opened/copied because of an I/O error" or similar.

(Note: What can you do then? You can't directly force the drive to swap that
block because it will only do that while writing. The easiest way would be
to simply delete the damaged file and either be sorry or copy it back from a
backup. This will leave the bad block as "available" again. One day later,
something tries to write on that block, sooner or later. At this time it
gets replaced.)

Now to your recent problem:

Someone suggested that you may have a bad block "close to the boot blocks"
of your directory structure.

If you simply format your drive, it will *not* wipe the entire disk, all
that it really does is to write out a new "empty directory". If the bad
block would have been there, it would get substituted and the drive works
again. However, let's assume that the bad block is within the reserved
directory area but none of the blocks that get written for the "empty
directory".

Then DU would rattle for some 30 seconds writing the fresh directory and
then follow up with some diagnostics if the drive appears valid. During that
process it would read/scan the entire directory area, stumple upon the bad
block while *reading*, get an error, and abort the whole process.

Bingo! That would explain your report.

Now you try to format the drive in a different way using Windows. This
different way uses different reserved areas, will not access the bad block
and works. Bingo!

Then you try to format it back on the Mac, it will access the original first
area again, stumble across the bad block again and fail again. Bingo!

As you see, that was a pretty good idea and it sounded very reasonable.

BUT - in the meantime you did this: You formatted the drive "low level",
meaning that the utility wrote zeros on *every* block on the entire drive
before doing anything else. If there would have been this single bad block
at the beginning, it would have been re-allocated at this stage and the
whole drive would be entirely OK afterwards.

Sadly your problems came back afterwards - so although the idea with that
block was very promising at the beginning, you have just disproved it - and
this means that the drive is probably dying.

If still under warranty, I would re-zero it one more time and ask LaCie to
get a new one.

From: MartinC on 26 Jul 2010 04:29
Nick Naym wrote:

>> Then you try to format it back on the Mac, it will access the original first
>> area again,
>
> It returns to that same area?? Why? It's no longer a "Mac" volume.

But you turn it into one (again). I got the impression that the drive works
fine as long as it is "Windows"-formatted and only throws errors once it is
formatted back to Mac. (Maybe I misunderstood this)

> Well, it seem that I can re-zero the drive, and reformat/reinitialize it.
> But then it only stays that way for a short while.

Then it is definitely damaged. If it would have been only a single bad block
on a very unlucky spot it would have been replaced at your zero-run.

If the drive behaves good for a short time afterwards but starts trouble
again, then it could be two/three possibilities:

1) The number of hidden blocks is (of course) limited. Once they are all
replaced the drive can't do it any longer and will start to throw errors
instantly even while writing. This means that a very large area is damaged
and that means the drive is dead.

2) It's not a bad block on the surface at all but something that the drive
itself damages while running. Either way, the drive is dead.

3) Of course it could be the interface/controller/internal wires. Then: dead
once again.

> The only times it seems to allow me to do anything again is after I shut down
> the Mac, power down the drive, and leave it alone for a long time. It may just
> be a coincidence, but it's been only after long periods of powered-down "rest"
> that it then seems to "recover" enough to allow me to erase it and/or
> reformat/reinitialize it.

This is possible, it means that the drive starts to fail once its
temperature rises. As stated above, you have already disproved the "one bad
block" theory by your zero-run that didn't solve it.

> Right now, it refuses to allow me to do anything. _Assuming_ it "recovers"
> enough to allow me to do _anything_, should I attempt to
> reformat/reinitialize, or simply go to DU's Erase tab, select one of the
> "Security Options..." (is the "Zero Out" sufficient?), and hit the
> "Erase..." button?

You're still under warranty, right? Then get it exchanged. Write them a note
that you formatted it writing zeros and that it still fails. That's
sufficient, there's nothing more you can/should/have to try in order to file
a complaint.

If you feel worried about your data then *zero* it once more before sending
it back.

> (is the "Zero Out" sufficient?)

This question may likely trigger a very heated debate... but to cut a long
story short: Yes, a one-pass zero-run is sufficient.

You will find stories on the internet that the secret services of the world
and evil hackers have ways to scan the drive and find shallow shades of
magnet information that allows for a restore...

<controversial advice>

Well... I put it this way: If it would be easy (or possible at all) to do
this then drive companies would simply sell drives with double capacity by
storing two different bytes on the same spot: The newly written one *and*
the one that used to be underneath.

I read a report that it *may* be possible to guess every 50th bit on the
surface while the remaining 49 will be random. Imagine a text where 49
letters are garbled and every 50th letter is correct (but always in random
order, you don't know which). Can you still read *anything* or just figure
out what it is?

The governments are very careful so there are strict rules that drives with
dangerous content *must* be formatted 7, 77, 777 times (whatever) and Apple
& Co. *must* provide this service in order to sell computers to them. So
this is why it is in, but outside of the X-Files series everyone really
knows that *one* zero-run really does the job ;-)

</controversial advice>

If you *don't* have warranty any more, please tell me - then you could try
something more.

From: MartinC on 26 Jul 2010 09:26
Nick Naym wrote:

> The drive just mounted after being shut down all night. It mounted as a
> fully formatted drive (i.e., it now appears just the way it did after the
> last time I reinitialized it).
>
> I can "zero out" the _drive_ two ways: via DU's Erase tab, or via DU's
> Partition tab; and I can "zero out" the _volume_ one way: via DU's Erase
> tab.
>
> I don't know which of these 3 ways is preferred because I don't know what
> the differences are.

The one that takes hours... (are you sure that you really did it that way?)

Click on the "Erase" tab, then you will see a button called "security
options" below (may be slightly different, I'm non-english here :-)

You should get a dialog with 4 options:
- don't delete data
- write zeroes
- write zeroes, 7 passes
- write zeroes, 35 passes

The second one (write zeroes, 1 pass) is the one to go. Depending on the
size of the drive, it will take some serious time. If you haven't done this
(i.e. if the dialog has a "don't delete" setting and your successful format
was pretty quick) then you may still have a possible bad block that may have
caused all this.

All formatting that is finished within some 1-2 minutes is nothing but
writing a small "empty" directory and leave everything else as it is. If you
do this, a new owner of your drive (or the company you send it to) could
easily recover most of your personal data.

> It's still under warranty...but I'm curious what "something more" you have
> in mind.

LaCie will not "repair" it, at least not for an acceptable price, so without
warranty you can throw it away... before you do that you could open the case
(this will *kill* your warranty, so don't do it for fun) and get out the
plain drive. If you connect it to an internal S-ATA bus then you can read
out the SMART data. This is like a black-box in an aeroplane, it will tell
you exactly what happened during the errors.

If it says OK then you know that the case/interface was broken and you have
a spare internal disk for future use. If the drive was broken, you will know
for sure.

Sadly there is *no* way to retrieve SMART through either USB or FW.

From: MartinC on 26 Jul 2010 11:03
Michelle Steiner wrote:

> Obligatory question: Hex zeroes or binary zeroes?

Hey... another invitation to a math/philosophical debate ;-)

It definitely is binary zeroes and *not* ASCII zeroes (= 48 or 0x30)

In math a hex 0x00 has exactly the same value as a decimal 0 and a binary 0
which is... well, just zero as in "one minus one"! ;-)

 |  Next  |  Last
Pages: 1 2
Prev: Q about minimized documents and Dock in Snow Leopard
Next: Mac - Windows filesharing search issue