bug#6131: [PATCH]: fiemap support for efficient sparse file copy

[Jim Meyering]

jeff.liu wrote:
...
> Sorry for the lack of detailed info for this point, except for removing the 
> fiemap->fm_start from
> the loop, I need to remove "fiemap->fm_start = (fm_ext[i-1].fe_logical + 
> fm_ext[i-1].fe_length);"
> out of the 'for (i = 0; i < fiemap->fm_mapped_extents; i++)" as well.
> So, if there is only one extent, at least 'i == 1' when the loop finished, 
> we'll not hit the
> 'fm_ext[-1]' issue.
>
> my thoughts of the fix looks like below:
>
> memset (fiemap, 0, sizeof fiemap_buf);
> do
>   {
>     ioctl (...);
>
>     for (i = 0; i < fiemap->fm_mapped_extents; i++)
>       {
>         ...
>       }
>     fiemap->fm_start = (fm_ext[i-1].fe_logical + fm_ext[i-1].fe_length);
>   } while (! last);

That is better.
Equivalent semantics to my change, but yours avoids unnecessarily
updating fiemap->fm_start for each iteration of the for loop.

...
>> For reference, here's what filefrag -v output looks like,
>> given a file with a nontrivial list of extents:
>>
>>   $ perl -e 'BEGIN{$n=16*1024; *F=*STDOUT}' \
>>          -e 'for (1..5) { sysseek(*F,$n,1)' \
>>          -e '&& syswrite *F,"."x$n or die "$!"}' > j
>>   $ filefrag -v j
>>   Filesystem type is: ef53
>>   File size of j is 163840 (40 blocks, blocksize 4096)
>>    ext logical physical expected length flags
>>      0       4  6258884               4
>>      1      12  6258892  6258887      4
>>      2      20  6258900  6258895      4
>>      3      28  6258908  6258903      4
>>      4      36  6258916  6258911      4 eof
>>   j: 6 extents found
> Do we need another test script for this test if we choose `filefrag' to 
> examine the extent info?

Yes, that's why I took the time to do the above.
I've already written most of it.  Will post shortly.