Re: [Qemu-devel] [PATCH v12 03/14] qcow2: Optimize bdrv_make_empty()

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2178 bytes --]

On 08/26/2014 03:36 PM, Max Reitz wrote:
> bdrv_make_empty() is currently only called if the current image
> represents an external snapshot that has been committed to its base
> image; it is therefore unlikely to have internal snapshots. In this
> case, bdrv_make_empty() can be greatly sped up by emptying the L1 and
> refcount table (while having the dirty flag set) and creating a trivial
> refcount structure.
> 
> If there are snapshots, fall back to the simple implementation (discard
> all clusters).
> 
> Signed-off-by: Max Reitz <mreitz@redhat.com>
> ---
>  block/blkdebug.c      |   2 +
>  block/qcow2.c         | 137 ++++++++++++++++++++++++++++++++++++++++++++------
>  include/block/block.h |   2 +
>  3 files changed, 126 insertions(+), 15 deletions(-)
> 

> +    } else {
> +        int l1_clusters;
> +        int64_t offset;
> +        uint64_t *new_reftable;
> +        uint8_t l1_ofs_rt_ofs_cls[20]; /* L1 offset; RT offset and clusters */
> +        uint64_t rt_entry;
> +
> +        ret = qcow2_cache_empty(bs, s->l2_table_cache);
>          if (ret < 0) {
> -            break;
> +            return ret;
> +        }
> +
> +        ret = qcow2_cache_empty(bs, s->refcount_block_cache);
> +        if (ret < 0) {
> +            return ret;
> +        }
> +
> +        /* Refcounts will be broken utterly */
> +        ret = qcow2_mark_dirty(bs);

qcow2_mark_dirty does assert(s->qcow_version >= 3).  But this code can
be reached when committing a qcow2 0.10 compat-level file, right?

> +        /* "Create" an empty reftable (one cluster) directly after the image
> +         * header and an empty L1 table three clusters after the image header;
> +         * the cluster between those two will be used as the first refblock */
> +        cpu_to_be64w((uint64_t *)&l1_ofs_rt_ofs_cls[ 0], 3 * s->cluster_size);
> +        cpu_to_be64w((uint64_t *)&l1_ofs_rt_ofs_cls[ 8],     s->cluster_size);
> +        cpu_to_be32w((uint32_t *)&l1_ofs_rt_ofs_cls[16], 1);

The offsets here feel a bit magic.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 539 bytes --]