Re: [Bug-ddrescue] Questions regarding copying algorithm

[Antonio Diaz Diaz]

Hello Bret.

Bret Quigley wrote:
> Firstly, thank you so much for creating such a remarkably effective tool. I
> have a friend who has a bad habit of dropping laptops, and ddrescue has
> helped a ton in the past.

Glad to know that ddrescue helped you.


> Apologies that this is not a bug report, but simply a few questions
> regarding the phase one "copying" algorithm. I'm sure that the information
> is right in front of me in the source code, but I'm unfortunately not
> familiar with C++.

I hope some day all the info will be in the manual. I try to clarify the 
manual every time I receive questions like yours.


> > From the manual, the copying phase "read[s] the non-tried parts of the
> input file, marking the failed blocks as non-trimmed and skipping beyond
> them, until all the rescue domain is tried."
>
> My reading of this is that ddrescue tries to copy the number of sectors
> specified in --cluster-size as a block, and that the block is marked as
> non-trimmed (*) or finished (+) in the logfile depending on the result.
> Why, then, do I see some blocks marked as bad-sector (-) during the copying
> phase? Shouldn't that not occur until the trimming phase, when the sectors
> are read individually?

The copying phase also tries to read individual sectors after skipping 
over a bad block. These sectors, if failed, are marked as bad-sector. 
See lines 137-138 of rescuebook.cc:

      const int alignment = skip_size ? hardbs() : softbs();
      Block b( pos, alignment );


> Secondly, I was wondering what relationship, if any, exists between
> --cluster-size and --skip-size. If ddrescue tries and fails to copy a
> block, is the next position calculated from the beginning or end of the
> block that just failed?

There is no relation between them. The size to skip is calculated from 
the end of the block that just failed.


> Thirdly, the manual states that "slow areas are also skipped and tried
> later in additional passes (before trimming)." I was wondering how this is
> reflected in the logfile. In other words, if a block is copied slowly, and
> ddrescue skips ahead as a result, are the skipped blocks simply left as
> non-tried?

Yes, because the skipped blocks have not been tried and may be that they 
are not slow at all.


Regards,
Antonio.