Re: [Qemu-devel] [PATCH 5/7] block: use fallocate(FALLOC_FL_PUNCH_HOLE) & fallocate(0) to write zeroes

[Max Reitz <mreitz@redhat.com>]

On 2015-01-27 at 08:51, Denis V. Lunev wrote:
> This sequence works efficiently if FALLOC_FL_ZERO_RANGE is not supported.
> Unfortunately, FALLOC_FL_ZERO_RANGE is supported on really modern systems
> and only for a couple of filesystems. FALLOC_FL_PUNCH_HOLE is much more
> mature.
>
> The sequence of 2 operations FALLOC_FL_PUNCH_HOLE and 0 is necessary due
> to the following reasons:
> - FALLOC_FL_PUNCH_HOLE creates a hole in the file, the file becomes
>    sparse. In order to retain original functionality we must allocate
>    disk space afterwards. This is done using fallocate(0) call
> - fallocate(0) without preceeding FALLOC_FL_PUNCH_HOLE will do nothing
>    if called above already allocated areas of the file, i.e. the content
>    will not be zeroed
>
> This should increase the performance a bit for not-so-modern kernels.
>
> Signed-off-by: Denis V. Lunev <den@openvz.org>
> CC: Kevin Wolf <kwolf@redhat.com>
> CC: Stefan Hajnoczi <stefanha@redhat.com>
> CC: Peter Lieven <pl@kamp.de>
> CC: Fam Zheng <famz@redhat.com>
> ---
>   block/raw-posix.c | 14 ++++++++++++++
>   1 file changed, 14 insertions(+)
>
> diff --git a/block/raw-posix.c b/block/raw-posix.c
> index 3c35b2f..c039bef 100644
> --- a/block/raw-posix.c
> +++ b/block/raw-posix.c
> @@ -967,6 +967,20 @@ static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData *aiocb)
>       }
>   #endif
>   
> +#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
> +    if (s->has_discard) {
> +        ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
> +                           aiocb->aio_offset, aiocb->aio_nbytes);
> +        if (ret < 0) {
> +            if (ret == -ENOTSUP) {
> +                s->has_discard = false;
> +            }
> +            return ret;
> +        }
> +        return do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes);
> +    }
> +#endif
> +

Sharing "has_discard" with handle_aiocb_discard() looks fine to me, 
because it's used for the the same do_fallocate() call there.

Once again, you should not abort if the first do_fallocate() returns 
ENOTSUP, because this is inconsistent with the behavior on the second 
call to handle_aiocb_write_zeroes() (where it falls through due to 
has_discard being false). Once again, this doesn't make a difference 
now, but very well might after the next patch.

And finally, do we need another has_foo for the fallocate(0) call? (like 
just "has_fallocate")

Max