Re: [Bug-ddrescue] Suggestion / feature request - bad head mapping

[Scott Dwyer]

Part 2:
> The third copy pass reads all the leftover non-tried blocks, from the 
> beginning, in order, with no skipping. This phase still reads in 
> blocks (128 sectors or 65536 bytes by default). The reason the copy is 
> done like this with three passes is that the first two passes are 
> capable of getting a big chunk of good data quickly, while the third 
> pass deals with the leftovers. And the leftovers are located in and 
> around the bad areas of the disk, which is what we were trying to 
> avoid on the first two passes.
>
> The next phase after the copy phase is the trimming phase. This phase 
> reads all the non-trimmed blocks in order. To be more accurate, it 
> reads one sector at a time from the start of each block until it 
> encounters an error, then reads from the end of that block until it 
> encounters another error. Meaning it reads the outsides of each block 
> until it hits a read error. If there is untried data left in the 
> middle it is marked as non-scraped. Depending on the nature of the 
> errors, this phase can take a long time.
>
> After the trimming phase comes the scraping phase (used to be the 
> splitting phase). The scraping phase reads all the non-scraped blocks 
> in order, one sector at a time from the beginning of each block. It 
> continues to do this until it reaches the end, at which point all the 
> sectors on the disk have been attempted. Depending on the nature of 
> the errors, this phase can take a very long time.
>
> After the trimming phase is done and if you have only a small error 
> size, you may wish to attempt retries of the bad area by using the 
> --retry-passes option (-r). However, just understand that by doing 
> this, you are focusing on a bad spot and it could make it worse (which 
> is why we save this for last). You also may not get any more 
> successful reads so don’t expect any miracles with retries. Also, if 
> you have a large error size, the retries will take a very long time, 
> possibly close to as long as the whole recovery took so far.
>
> Ddrescue version 1.19 has some very good options for controlling what 
> phases and copy passes you wish to do. For instance, you can skip the 
> trimming phase with the --no-trim option. You can skip the scraping 
> phase with the --no-scrap options. Note that if you use the no-trim 
> option, the scraping phase will still try to run, but because there 
> was no trimming there should be no areas marked as non-scraped so the 
> scraping phase will not do anything. To continue on with the rescue 
> including these phases just remove the options from your ddrescue command.
>
> The copy passes can also be controlled with the --cpass option. So if 
> you wanted to just do copy pass 1 and 2 you would use the options 
> --cpass=1,2 --no-trim --no-scrape. This will stop the rescue after the 
> first two copy passes so you can examine it and determine the best way 
> to continue. Just be aware of something very important. If you don’t 
> run copy pass 3, you will most likely have non-tried areas left over! 
> Also, if copy pass 3 has not completed (either by options or you stop 
> the rescue with ctrl-c) and then you run the ddrescue command again, 
> it will start with copy pass 1 again.
>
> The following assumes a default ddrescue command (no special pass 
> options). So if you stop ddrescue during copy phase 1 and then resume, 
> it will continue from where it left off. If you stop ddrescue during 
> copy phase 2 or 3 it will start with copy pass 1 again, but from the 
> last input position.
>
> So now that you know this, if you only perform the first two copy 
> passes (--cpass=1,2 --no-trim --no-scrape), when you resume you may 
> wish to use the option --cpass=3 so it will do what it should next.
>
> I intend my next post to be about how to implement the commands, but 
> it may be a bit before I can get to it.