On Sat, Jun 4, 2011 at 1:44 AM, Fam Zheng <famc...@gmail.com> wrote:
> The `qemu-img info` results for mono flat image are no longer
> accurate, as the "disk size" was the length of bs->file, which is not
> the case for multi file images (such as vmdk images with multiple
> files).
> The new field disk_size in BlockDriverState can be used by block
> driver to tell the exact disk size of block image (only valid if the
> field is set to non-zero).
>
> Signed-off-by: Fam Zheng <famc...@gmail.com>
> ---
> block/vmdk.c | 1 +
> block_int.h | 1 +
> qemu-img.c | 6 +++++-
> 3 files changed, 7 insertions(+), 1 deletions(-)
Instead of introducing a variable please add a new BlockDriver
.bdrv_get_allocated_file_size() function pointer and move
get_allocated_file_size() from qemu-img.c into block/raw-posix.c and
block/raw-win32.c.
qemu-img.c should call bdrv_get_allocated_file_size() and
bdrv_get_allocated_file_size() could look like this:
int bdrv_get_allocated_file_size(BlockDriverState *bs, int64_t *size)
{
BlockDriver *drv = bs->drv;
if (!drv) {
return -ENOMEDIUM;
}
if (drv->bdrv_get_allocated_file_size) {
return drv->bdrv_get_allocated_file_size(bs, size);
}
if (bs->file) {
return bdrv_get_allocated_file_size(bs->file, size);
}
return -ENOTSUP;
}
> @@ -607,6 +607,7 @@ static int vmdk_open_desc_file(BlockDriverState
> *bs, int flags)
> extent = s->extents;
> vmdk_parse_extents(buf, s->extents, bs->file->filename);
> bs->total_sectors = extent->sectors;
> + bs->disk_size = bdrv_getlength(bs->file) + bdrv_getlength(extent->file);
This should call bdrv_get_allocated_file_size() so add together the
actually allocated amount, not what bdrv_getlength() returns.
Stefan
