Re: [dm-devel] [PATCH 1/5] block: Implement support for WRITE SAME

[Mike Snitzer <snitzer redhat com>]

On Wed, Feb 08 2012 at  6:12pm -0500,
Martin K. Petersen <martin petersen oracle com> wrote:

> >>>>> "Mike" == Mike Snitzer <snitzer redhat com> writes:
> 
> >> + BUG_ON(length != bdev_logical_block_size(bdev));
> 
> Mike> Why require @length to be passed in if we can easily determine the
> Mike> only correct length, via bdev_logical_block_size, locally?
> 
> Just sanity checking since the caller is passing in a buffer pointer.
> See my comment in the patch series introduction about a page vs a
> buffer. I don't really care whether we pick one or the other. Whatever
> makes the most sense to dm, md and the filesystems...

OK, I see -- might make sense to use a page.  But I don't have a
strong opinion (and not sure I will have one in the near-term).

I wrote a patch for dm-kcopyd (and dm-io) to add support for WRITE_SAME,
but it doesn't use the new blkdev_* interfaces you've provided -- only
because I needed to wire it up to dm-thinp which uses dm_kcopyd_zero()
to provide async zeroing.

So the patch open codes a small portion of bio prep code from
blkdev_issue_write_same() in dm-io's do_region().  I'll post it once I
verify it works.

Mike