Re: [Qemu-devel] [PATCH 1/1] block: change default memory alignment for block requests to 4096

[Denis V. Lunev]

On 28/01/15 23:07, Paolo Bonzini wrote:
> On 28/01/2015 19:49, Denis V. Lunev wrote:
> > The following sequence
> >      int fd = open(argv[1], O_RDWR | O_CREAT | O_DIRECT, 0644);
> >      for (i = 0; i < 100000; i++)
> >              write(fd, buf, 4096);
> > performs 10% better if buf is aligned to 4096 bytes rather then to
> > 512 bytes on HDD with 512/4096 logical/physical sector size.
> >
> > The difference is quite reliable.
> The 10% difference, however, is probably not enough to cover the cost of
> providing a bounce buffer if a guest is (rightfully) using a 512-byte
> aligned buffer: bs->bl.opt_mem_alignment is in fact badly named and it
> should be bs->bl.min_mem_alignment instead.
>
> Instead, you probably should patch bdrv_opt_mem_align to return at least
> 4096, and leave the detection logic intact.  This will let
> qemu_blockalign return a properly aligned buffer to qemu-img and other
> in-process allocations, without negatively affecting the guest.
>
> Thanks,
>
> Paolo
ok, this looks good to me :)


> > Signed-off-by: Denis V. Lunev <address@hidden>
> > CC: Kevin Wolf <address@hidden>
> > CC: Stefan Hajnoczi <address@hidden>
> > ---
> >   block.c           | 4 ++--
> >   block/raw-posix.c | 4 ++--
> >   2 files changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/block.c b/block.c
> > index d45e4dd..bc5d1e7 100644
> > --- a/block.c
> > +++ b/block.c
> > @@ -543,7 +543,7 @@ void bdrv_refresh_limits(BlockDriverState *bs, Error **errp)
> >           bs->bl.max_transfer_length = bs->file->bl.max_transfer_length;
> >           bs->bl.opt_mem_alignment = bs->file->bl.opt_mem_alignment;
> >       } else {
> > -        bs->bl.opt_mem_alignment = 512;
> > +        bs->bl.opt_mem_alignment = 4096;
> >       }
> >   
> >       if (bs->backing_hd) {
> > @@ -966,7 +966,7 @@ static int bdrv_open_common(BlockDriverState *bs, 
> > BlockDriverState *file,
> >   
> >       bs->open_flags = flags;
> >       bs->guest_block_size = 512;
> > -    bs->request_alignment = 512;
> > +    bs->request_alignment = 4096;
> >       bs->zero_beyond_eof = true;
> >       open_flags = bdrv_open_flags(bs, flags);
> >       bs->read_only = !(open_flags & BDRV_O_RDWR);
> > diff --git a/block/raw-posix.c b/block/raw-posix.c
> > index ec38fee..d1b3388 100644
> > --- a/block/raw-posix.c
> > +++ b/block/raw-posix.c
> > @@ -266,7 +266,7 @@ static void raw_probe_alignment(BlockDriverState *bs, int 
> > fd, Error **errp)
> >       if (!s->buf_align) {
> >           size_t align;
> >           buf = qemu_memalign(MAX_BLOCKSIZE, 2 * MAX_BLOCKSIZE);
> > -        for (align = 512; align <= MAX_BLOCKSIZE; align <<= 1) {
> > +        for (align = 4096; align <= MAX_BLOCKSIZE; align <<= 1) {
> >               if (pread(fd, buf + align, MAX_BLOCKSIZE, 0) >= 0) {
> >                   s->buf_align = align;
> >                   break;
> > @@ -278,7 +278,7 @@ static void raw_probe_alignment(BlockDriverState *bs, int 
> > fd, Error **errp)
> >       if (!bs->request_alignment) {
> >           size_t align;
> >           buf = qemu_memalign(s->buf_align, MAX_BLOCKSIZE);
> > -        for (align = 512; align <= MAX_BLOCKSIZE; align <<= 1) {
> > +        for (align = 4096; align <= MAX_BLOCKSIZE; align <<= 1) {
> >               if (pread(fd, buf, align, 0) >= 0) {
> >                   bs->request_alignment = align;
> >                   break;