Re: [Bug-ddrescue] Question & suggestion

[Antonio Diaz Diaz]

Hello David,

address@hidden wrote:
> I am trying to rescue a drive (120GB) which has large blocks of errors then
> blocks of readable data. What I want to do is to try and read a large block
> (1MB) of data, if it fails then skip on by 1MB and try again.  So I launched
> ddrescue with -c 1024 (which should read 512K chunks as I understand it).
> But I am not seeing what I expect in syslog, I would expect to see failed to
> read sector @ n, followed by failed to read sector @ n+1024 in the error
> areas, but the sector numbers go up by varying amounts, sometimes only a few
> 10's of sectors. It showing 40+minutes since last successful read so it
> doesn't seem to be reading any good data (or is that the last time an entire
> block was read without error?).

The amount skipped by ddrescue on a read error (or on a slow read) is 
quite variable. It depends on the previous history of read successes and 
failures.

On the first read error, ddrescue skips 'skipbs' bytes (currently 
defined as "max( 65536, 16 * hardbs )"). Then it reduces the skip_size 
every good read and doubles it every failed read.

A change that could achieve more or less what you want without modifying 
the default behaviour would be to redefine 'skipbs' as "max( 65536, 
softbs )"

BTW, the "x minutes since last successful read" means since the last 
time ddrescue managed to read something, even if it was 1 single byte.


> Suggestion
> Where a source drive goes offline during recovery is it possible to identify
> this from the returned error code and abort ddrescue, rather than filling
> the log up with failed sectors?  EG on IDE I got DID_Invalid & on usb I get
> drive offline.  I think the logic should be 'media error code' carry on,
> device / drive error 'abort'...

DDrescue is device-agnostic and does not even try to access low-level 
device error codes.

Ddrescue already stops if the drive dissapears from /dev (see example 7 
in the manual), but if the drive does not dissapear, the only way of 
stopping it in case of total drive error is with --max-error-rate. See 
also the --try-again option.


Regards,
Antonio.