Re: [Qemu-block] [PATCH 0/6] block: Enable byte granularity I/O

[Kevin Wolf]

Am 13.06.2016 um 15:27 hat Stefan Hajnoczi geschrieben:
> On Wed, Jun 08, 2016 at 04:10:05PM +0200, Kevin Wolf wrote:
> > Previous series have already converted some block drivers to byte-based 
> > rather
> > than sector-based interfaces. However, the common I/O path as well as 
> > raw-posix
> > still enforced a minimum alignment of 512 bytes because some sector-based 
> > logic
> > was involved.
> > 
> > This patch series removes these limitations and a sub-sector request 
> > actually
> > ends up as a sub-sector syscall now.
> > 
> > Kevin Wolf (6):
> >   block: Byte-based bdrv_co_do_copy_on_readv()
> >   block: Prepare bdrv_aligned_preadv() for byte-aligned requests
> >   block: Prepare bdrv_aligned_pwritev() for byte-aligned requests
> >   raw-posix: Switch to bdrv_co_* interfaces
> >   raw-posix: Implement .bdrv_co_preadv/pwritev
> >   block: Don't enforce 512 byte minimum alignment
> > 
> >  block.c               |   2 +-
> >  block/io.c            | 125 
> > +++++++++++++++++++++++++-------------------------
> >  block/linux-aio.c     |  83 ++++++++++++++++++++++++---------
> >  block/mirror.c        |  10 ++--
> >  block/raw-aio.h       |   2 +
> >  block/raw-posix.c     |  61 ++++++++++++------------
> >  include/block/block.h |  10 ++--
> >  7 files changed, 169 insertions(+), 124 deletions(-)
> 
> I've taken an initial look and it looks good.  Will review next revision
> in depth so it can be merged after Eric's comments have been addressed.

Eric commented a lot, but only requested very few minor changes that
wouldn't strictly require resending the series. If you think it's
worthwhile to send a v2 for them, I can do that, but it shouldn't make a
big difference for your review.

Kevin

pgpylCdXt4hkg.pgp
Description: PGP signature