On 12/03/2014 06:37 AM, Max Reitz wrote:
> update_refcount() and qcow2_update_cluster_refcount() currently take a
> signed addend. At least one caller passes a value directly derived from
> an absolute refcount that should be reached ("l2_refcount - 1" in
> expand_zero_clusters_in_l1()). Therefore, the addend should be unsigned
> because unsigned overflow is well-defined in contrast to signed
> overflow. This will be especially important for 64 bit refcounts.
>
> Because update_refcount() then no longer knows whether the refcount
> should be increased or decreased (which is important for setting the
> refblock-L2-table cache dependency and for overflow/underflow checks),
> it now requires an additional flag which specified exactly that. The
> same applies to qcow2_update_cluster_refcount().
>
> Signed-off-by: Max Reitz <[email protected]>
> ---
> block/qcow2-cluster.c | 2 +-
> block/qcow2-refcount.c | 63
> ++++++++++++++++++++++++++++++++------------------
> block/qcow2.h | 3 ++-
> 3 files changed, 44 insertions(+), 24 deletions(-)
>
> @@ -1677,7 +1695,8 @@ static void compare_refcounts(BlockDriverState *bs,
> BdrvCheckResult *res,
>
> if (num_fixed) {
> ret = update_refcount(bs, i << s->cluster_bits, 1,
> - (int)refcount2 - (int)refcount1,
> + refcount2 - refcount1,
> + refcount1 > refcount2,
> QCOW2_DISCARD_ALWAYS);
Nice - my comment on 4/26 is indeed addressed here (so I was correct in
guessing that the series touched that area again, and no need to respin).
Reviewed-by: Eric Blake <[email protected]>
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
