Re: [PATCH v6 04/10] qcow2: Support BDRV_REQ_ZERO_WRITE for truncate

[Kevin Wolf <kwolf@redhat.com>]

Am 23.04.2020 um 17:18 hat Eric Blake geschrieben:
> On 4/23/20 10:01 AM, Kevin Wolf wrote:
> > If BDRV_REQ_ZERO_WRITE is set and we're extending the image, calling
> > qcow2_cluster_zeroize() with flags=0 does the right thing: It doesn't
> > undo any previous preallocation, but just adds the zero flag to all
> > relevant L2 entries. If an external data file is in use, a write_zeroes
> > request to the data file is made instead.
> > 
> > Signed-off-by: Kevin Wolf <kwolf@redhat.com>
> > Reviewed-by: Max Reitz <mreitz@redhat.com>
> > ---
> >   block/qcow2-cluster.c |  2 +-
> >   block/qcow2.c         | 33 +++++++++++++++++++++++++++++++++
> >   2 files changed, 34 insertions(+), 1 deletion(-)
> > 
> 
> > +    if ((flags & BDRV_REQ_ZERO_WRITE) && offset > old_length) {
> > +        uint64_t zero_start = QEMU_ALIGN_UP(old_length, s->cluster_size);
> > +
> > +        /*
> > +         * Use zero clusters as much as we can. qcow2_cluster_zeroize()
> > +         * requires a cluster-aligned start. The end may be unaligned if it is
> > +         * at the end of the image (which it is here).
> > +         */
> > +        ret = qcow2_cluster_zeroize(bs, zero_start, offset - zero_start, 0);
> > +        if (ret < 0) {
> > +            error_setg_errno(errp, -ret, "Failed to zero out new clusters");
> > +            goto fail;
> > +        }
> > +
> > +        /* Write explicit zeros for the unaligned head */
> > +        if (zero_start > old_length) {
> > +            uint8_t *buf = qemu_blockalign0(bs, s->cluster_size);
> > +            QEMUIOVector qiov;
> > +            qemu_iovec_init_buf(&qiov, buf, zero_start - old_length);
> > +
> > +            qemu_co_mutex_unlock(&s->lock);
> > +            ret = qcow2_co_pwritev_part(bs, old_length, qiov.size, &qiov, 0, 0);
> 
> This works, but would it be any more efficient to use
> qcow2_co_pwrite_zeroes?  If the head of the cluster is already zero, then
> qcow2_co_pwrite_zeroes can turn into qcow2_cluster_zeroize for this cluster,
> while qcow2_co_pwritev_part cannot.

The problem is that qcow2_co_pwrite_zeroes() will return -ENOTSUP if the
request is still unaligned after this optimisation. I would have to go
through the generic bdrv_co_pwrite_zeroes() to get the fallback, but I
can't do that because bs->total_sectors isn't updated yet.

Kevin