On Wed, Jun 01, 2022 at 10:12:26AM -0600, Keith Busch wrote:

On Wed, Jun 01, 2022 at 08:28:31AM -0600, Keith Busch wrote:

quoted

On Wed, Jun 01, 2022 at 12:11:40AM -0700, Eric Biggers wrote:

quoted

I still don't think you've taken care of all the assumptions that bv_len is a
multiple of logical block size, or at least SECTOR_SIZE.  Try this:

	git grep -E 'bv_len (>>|/)'

There are only 8 drivers that set the request_queue's dma alignment, which are
the only ones that could be affected from this patch series.

It's actually even simpler to audit than that. Of the 8 drivers that explicitly
set dma alignment, only 3 set it to something smaller than a sector size. None
of them assume any particular bv_len, so I think we're fine.

Eric,

Do you have any more concerns on this area? I'm reasonably confident this is
safe with all the existing users, and I have a new series ready to go with the
trivial fix-ups from the last round. I don't want to post it, though, if you
think I've missed something.

Thanks,
Keith