On Thu, Mar 30, 2017 at 07:15:50PM -0400, Mike Snitzer wrote: > I got pretty far along with implementing the DM thinp support for > WRITE_ZEROES in terms of thinp's DISCARD support (more of an > implementation detail.. or so I thought). > > But while discussing this effort with Jeff Moyer he asked: shouldn't the > zeroed blocks be provisioned? This is a fairly embarassing question not > to be able to answer in the moment. So I clearly need to learn what the > overall intent of WRITE_ZEROES actually is. It is to ensure the that the block range it's called on returns zeroes on future reads. Currently if REQ_UNMAP is set you may free the space allocation, else not. After looking at the callers I think this is the wrong way around, so I'm going to invert the flag, so that the two callers that care that blocks are allocated will set it instead.
