On Jul 30, 2014, at 11:18 AM, David Sterba [off-list ref] wrote:

Add a new member to fiemap_extent that represents the physical extent
length. This value is undefined if the flag EXTENT_PHYS_LENGTH is not
set.

The description here of PHYS_LENGTH makes sense...

The patch description should also mention the name of the new member,
namely "fe_phys_length"

+#define FIEMAP_EXTENT_PHYS_LENGTH	0x00000010 /* Physical length of extent
+						    * not the same as logical */

But the comment doesn't match.  This implies that if PHYS_LENGTH is
set, fe_phys_length != fe_logi_length, but I don't think that is
necessarily correct.  I think it makes more sense to just set
PHYS_LENGTH when fe_phys_length is valid, and if PHYS_LENGTH is not
set then fe_phys_length aware applications should just use
fe_phys_length = fe_logi_length, and older applications would just
use fe_length for both as they would already today.

Cheers, Andreas