[ILUG] Rescue file from damaged CDR?

[Kenn Humborg]

On Thu, Sep 14, 2000 at 08:00:32PM +0100, Conor Daly wrote:
> Ideas anyone?
> 
> Trying to rescue a ~28Mb file from a damaged CDR.  I'm trying 
> 
> dd if=/mnt/cdrom/file of=first.out
> 
> once I get an I/O error I get a count of blocks read by dd so I do it
> again using 
> 
> dd if=/mnt/cdrom/file of=seconf.out skip=<blocks already read>+1
> 
> and so on until I run out of blocks to read.  I'll then use a single block
> fragment to pad out to the original size and do a 
> 
> cat first.out block-fragment second.out block-fragment third.out ......  > recovered-file
> 
> Will this work?  Is there a better way? 

You might be able to automate this with the noerror and sync options to dd:

   dd if=/mnt/cdrom/file of=file bs=2048 noerror sync

'noerror' says continue after read errors.  'sync' says pad out all input
blocks with NULs.  

However, I don't know the failed read results in an input block with zero
length that gets padded out and written, or results in that block not
being written _at_all_ to the output file.

To test it, use count=<num-blocks> to restrict the copy to the start of
the file up to just past the first damaged block.  Then check if the
output file is num-blocks*2048, or smaller.  If smaller, this won't work.

Alternatively, do it piecemeal:

   #!/bin/sh

   INFILE=/mnt/cdrom/file
   OUTFILE=file

   BLOCK_SIZE=2048
   NUM_BLOCKS=14000   # 28MB/2048 - work this out exactly
                      # ...or better yet, work it out in code

   for i in `seq 0 $[ $NUM_BLOCKS - 1 ] ` ; do

      dd if=$INFILE of=$OUTFILE bs=$BLOCK_SIZE count=1 seek=$i skip=$i

   done

If doesn't matter if dd writes a block of zeros or not for a bad block
here, because the next block will still be written at the correct
position (skip=$i), extending the file if necessary.

Later,
Kenn