Re: [Bug-ddrescue] Sparse and preallocate questions/feature idea

[Scott Dwyer]

On 2/24/2014 12:35 PM, Zeniff Martineau wrote:
> General followup sparse/usage check question (not specific to ddrescue):
> I was thinking if estimating could be possible by testing a few chunks
> at a time progressively toward the end of the drive, skipping large
> amounts, to see when/where the drive seems to have stopped "filling
> up". On a healthy drive, would this make sense for a very rough usage
> guess? Or maybe the drive just puts data anywhere without regard to
> physical location, so that wouldn't work? Just curious, since I think
> I remember running photorec on a very large drive with almost no
> usage, and it seemed to finish pretty fast, so I was assuming it was
> able to check something like that. But I will ask them too; I just
> wondered about that for ddrescue too. :)
If you are trying to use ddrescue because of its ability to do sparse 
writes, and you are using it on good drives (NOT failing drives), then 
how about this for an idea.

Mount the drive on a computer with its native filesystem if not Linux 
(if the drive is NTFS, hook it up to a Windows computer, ect...). Run 
the normal disk clean up utilities available to empty trash/recycle bin 
and stuff. Then look for a utility that will erase (wipe) unused space 
on a drive. You will want to find one that will allow you to only do a 
single pass of 00's.

Another alternative is to write a large file of zero's, and then delete 
it. Although I am not sure I would recommend that option unless you are 
careful, as I have had bad results from the accidental and unwanted 
filling a drive, and then it would not let me delete anything. If you 
mount the filesystem in Linux, you could use dd or even ddrescue with 
/dev/zero as the source to create such a file, and then the rm command 
to remove the output file. Again, you may want to be careful by leaving 
yourself a bit of a cushion and not create a file(s) that completely 
fill the drive.

Once either method is done, then in theory, the used portion of the disk 
reported by the filesystem should be about the size it would take up 
when using sparse writes. Note that I said "in theory" and "should", you 
would have to test to be sure.

Scott