Re: [Bug-ddrescue] Reverse mode?

[Antonio Diaz Diaz]

Hello Théo,

Théo Koelz wrote:
> I understand that a reverse mode has been added in the latest ddrescue
> but the ddrescue manual isn't to clear about what it does (I thought
> ddrescue's algorithm automatically switched to reverse mode when
> needed). Can you explain us what it does and what is its use?

In my opinion, reverse mode is only useful in rare occasions.

Note that ddrescue *reads backwards* when it is needed (trimming skipped 
blocks). This is not the same that "switching to reverse mode when needed".

As you can see in the Algorithm section of the manual[1], ddrescue 
performs four main passes; copying, trimming, splitting and retrying. In 
"normal" (not reverse) mode, copying, splitting and retrying are run 
forwards, while trimming is run backwards from the end of each 
non-trimmed block.

Reverse mode inverts all the passes. This means that copying, splitting 
and retrying are run backwards, while trimming is run forwards from the 
beginning of each non-trimmed block.

Given that reading backwards is much slower than reading forwards, and 
that trimming is usually the pass which reads less data by a wide 
margin, reverse mode tends to be much slower than normal mode. Even so, 
it seems some users have found cases where reverse mode helped them to 
recover the data.

[1] http://www.gnu.org/software/ddrescue/manual/ddrescue_manual.html


Best regards,
Antonio.