Re: [Bug-ddrescue] Interpreting logfile of restore to get list of corrupted files

[Ariel]

On Tue, 26 Sep 2006, Chris Witham wrote:

> I have a hard drive that was failing, and I was referred[1] to
> ddrescue.  It appears to have done a great job, but it reported over
> 1600 errors.  I am trying to find a way to turn the 5kB logfile from
> the restore into a listing of affected files so I know what I need to
> replace.

http://smartmontools.sourceforge.net/BadBlockHowTo.txt

Skip the part about the LBA and partition. Start at the line beginning 
Second Step.

I don't have, or know about, a script to do this.

But:

debug2fs is semi-scriptable using -f, both icheck and ncheck are very 
slow, so it's faster to give the entire list of numbers in a single line 
for each command, then parse the output. (So it does just a single pass 
over the fs.)

Like: icheck 155 667 333 etc

Then parse the output and:
ncheck <number> <number2> etc

Keep in mind ncheck can return more then one file name per inode, so 
don't use a hash on the inode to store the result.

A few other notes:

ddrescue can use a different block size then 512 (using -b). So I suggest 
you normalize the log file from block to bytes, then convert back to 
blocks this time using the fs block size. i.e. (int)(d*b/B) (instead of 
the equation given there, which includes the LBA).

Where d is the ddrescue block number, b is the hardware block size used 
with -b, and B is the filesystem block size.

Also, do not test your code on the bad hard disk, since it's possible that 
the more you use it, the more damage can happen. Make a tiny throwaway fs 
to play with.

        -Ariel

PS. If you write a script be sure to email it, maybe Antonio will include 
it in the release package.