Re: [PATCH 3/4] ublk: add feature UBLK_F_AUTO_ZERO_COPY

[Ming Lei <ming.lei@redhat.com>]

On Sat, Apr 26, 2025 at 10:04:46PM -0600, Keith Busch wrote:
> On Sun, Apr 27, 2025 at 11:10:30AM +0800, Ming Lei wrote:
> > On Sat, Apr 26, 2025 at 08:34:40PM -0600, Keith Busch wrote:
> > > 
> > > This is very similiar to something I proposed off-list, and the feedback
> > 
> > Looks we both think of it, :-)
> 
> Yeah, for real. I was a bit dismayed when I learned of such use cases.
> So much simplicity and elegance went away...

That is reality, and probably these use cases may be addressed elegantly too
in future...

>  
> > > back then was this won't work because the back-end ring that wants to
> > > use the zero-copy buffer isn't the same as the ublk server ring
> > > recieving notification of a new command; the ublk driver has no idea
> > > which uring to register the bvec with. Also, this is using the request
> > > "tag" as the io_uring buf index, which wouldn't work when the ublk
> > > server ring handles multiple ublk devices due to the tag collisions.
> > > 
> > > If you're can make those trade-offs, then this is a great simplification
> > > to the whole thing.
> > 
> > The io_uring fd & buffer index can be provided from 'ublksrv_io_cmd'.
> > 
> > https://lore.kernel.org/linux-block/aA2RNG3-WzuQqEN6@fedora/
> > 
> > If we only support IORING_ENTER_REGISTERED_RING, 32bit is enough for
> > io_uring fd & buffer index, and there is still 64bits available if not
> > taking UBLK_F_ZONED into account.
> 
> We still need a registered sparse table for the backend ring. I think
> maybe a simple ida from the ublk driver to select an index may let the
> daemon register something reasonably small.

Yeah, I think it is reasonable to let userspace register the sparse table,
and we can document it in UAPI.


thanks,
Ming