Re: [Qemu-devel] [PATCH] qemu-img: return allocated size for block device with qcow2 format

[Max Reitz <mreitz@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 3077 bytes --]

Hi,

[replying to this version because the previous mail doesn't seem to have
made it to the mailing lists for whatever reason]

On 2018-05-02 15:34, Ivan Ren wrote:
> qemu-img info with a block device which has a qcow2 format always
> return 0 for disk size, and this can not reflect the qcow2 size
> and the used space of the block device. This patch return the
> allocated size of qcow2 as the disk size.

I'm not quite sure whether you really need this information for block
devices (I tend to agree with Eric that wr_highest_cluster is the more
important information there), but I can imagine it just being nice to have.

So the basic idea makes sense to me, but I think the implementation can
be simplified and the reporting in qemu-img should be done differently.

> 
> Signed-off-by: Ivan Ren <ivanren@tencent.com>
> ---
>  block/qcow2-bitmap.c |  69 +++++++++++++++++
>  block/qcow2.c        | 212 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  block/qcow2.h        |  42 ++++++++++
>  3 files changed, 323 insertions(+)

The whole implementation reminds me a lot of qcow2's check function,
which basically just recalculates the refcounts.  So I'm wondering
whether you could just count how many clusters with non-0 refcount there
are and thus simplify the implementation dramatically.

[...]

> +static int64_t qcow2_get_allocated_file_size(BlockDriverState *bs)
> +{
> +    struct stat st;
> +    if (stat(bs->filename, &st) < 0 || !S_ISBLK(st.st_mode)) {
> +        goto get_file_size;
> +    }

This definitely doesn't work because nobody guarantees that bs->filename
is something that stat() can work with.  I'm aware that you need to do
the S_ISBLK() check somewhere, but the qcow2 driver is not the right place.

I don't really have a good way around this, though.  These things come
to mind:

(1) We could let file-posix report an error for S_ISBLK because the
information is known to be usually useless -- but I think that is not
quite the right thing to do because maybe some block devices do report
useful information there, I don't know.

(2) Or we introduce a new field in qemu-img info (and thus in ImageInfo,
too, I suppose?).  qemu-img info (or rather bdrv_query_image_info())
could detect whether the format layer supports
bdrv_get_allocated_file_size, and if so, it emits that information
separately from the allocated size of bs->file->bs.  But that would
break vmdk...

(3) As a hack, qcow2_get_allocated_file_size() could first always call
bdrv_get_allocated_file_size(bs->file->bs), and if that returns 0 (which
is absolutely impossible for qcow2 files because they have an image
header that takes up some space), we fall back to
qcow2_get_block_allocated_size().  While I consider it a hack, I can't
come up with a scenario where it wouldn't work.

Max

> +
> +    return qcow2_get_block_allocated_size(bs);
> +
> +get_file_size:
> +    if (bs->file) {
> +        return bdrv_get_allocated_file_size(bs->file->bs);
> +    }
> +    return -ENOTSUP;
> +}
> +


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]