Re: Suggestion about error control

[Antonio Diaz Diaz]

Hello kickman.

Thanks for your message and sorry for the late answer. I have been very busy.

anonymous wrote:
> The two last can recognize and distinguish the error codes got from the
> kernel after every read /write request. For example, when hddsuperclone
> detects the disk is offline (not ready) - it just waits for it to become
> ready again. So DMDE does. It shows different error codes allowing the
> operator to make the right decision about what to do. But my favourite
> tool ddrescue never made such a difference to the error codes! Either the
> error code was "CRC error" or "Device not ready!"

I suppose those tools (hddsuperclone and DMDE) use non-portable ways to get 
those codes, because neither "CRC error" nor "Device not ready!" appear 
among the error codes returned by 'read'[1].

[1] http://pubs.opengroup.org/onlinepubs/9699919799/functions/read.html

Bypassing the kernel should not be done lightly on POSIX systems. As a 
portable program, ddrescue must limit itself to the error codes returned by 
'read' as documented by POSIX. As a first approximation, 'EIO' may 
correspond to 'CRC error' and 'ENXIO' seems the best match for 'Device not 
ready'.

I'll implement some form of differential response to error codes in the next 
version of ddrescue. Maybe ddrescue could stop or ask the user when 
readblockp sets errno to EBADF, ESPIPE, or ENXIO, for example.

It may be also possible to implement a non-portable way to retrieve the 
drive status, and compile it conditionally with '--enable-non-posix'.

Feedback is welcome.

Best regards,
Antonio.