Re: [Bug-ddrescue] Why doesn't ddrescue revert to full speed long after a bad read?

[Antonio Diaz Diaz]

Paul L Daniels wrote:
> Without having yet looked at the code, could it be a 1-off error in the
> fallback speed code in ddrescue?  I guess I should go check the source
> code and see how ddrescue decides when it increases the speed back.

When ddrescue finds errors it skips ahead. The more the errors, the more 
it skips (see 'skip_size' in rescuebook.cc). After good data is found 
again, skip_size is decreased by the amount of data read. But as long as 
skip_size remains greater than 0, reads are made sector by sector.

I guess that, for slow drives like mine, sector-sized reads are still 
faster than the drive can deliver data, so I do not notice any slowdown.

Maybe the current algorithm is too conservative for large skip sizes and 
skip_size should be zeroed after some number of consecutive good sectors 
have been found.

Another possibility is that you are using a --min-read-rate slower than 
what sector by sector reads can deliver. In this case, after the first 
error ddrescue can't regain speed (all reads are considered slow reads).


Regards,
Antonio.