On Mon, May 08, 2017 at 11:46:14PM -0700, Nicholas A. Bellinger wrote:

That said, simply propagating up q->limits.max_write_zeroes_sectors as
dev_attrib->unmap_zeroes_data following existing code still looks like
the right thing to do.

It is not.  Martin has decoupled write same/zeroes support from discard
support.  Any device will claim to support it initially, and we'll
only clear the flag if a Write Same command fails.

So even if LBPRZ is not set you can trivially get into a situation
where discard is supported through UNMAP, and you'll incorrectly
set LBPRZ and will cause data corruption.