Re: [Qemu-devel] [PATCH 3/3] qcow2: Support for fixing refcount inconsistencies

[Stefan Hajnoczi <stefanha@gmail.com>]

On Fri, May 11, 2012 at 5:48 PM, Kevin Wolf <kwolf@redhat.com> wrote:
> @@ -1205,9 +1206,31 @@ int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res)
>
>         refcount2 = refcount_table[i];
>         if (refcount1 != refcount2) {
> +
> +            /* Check if we're allowed to fix the mismatch */
> +            int *num_fixed = NULL;
> +            if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
> +                num_fixed = &res->leaks_fixed;
> +            } else if (refcount1 < refcount2 && (fix & BDRV_FIX_ERRORS)) {
> +                num_fixed = &res->corruptions_fixed;
> +            }
> +
>             fprintf(stderr, "%s cluster %d refcount=%d reference=%d\n",
> -                   refcount1 < refcount2 ? "ERROR" : "Leaked",
> +                   num_fixed != NULL     ? "Repairing" :
> +                   refcount1 < refcount2 ? "ERROR" :
> +                                           "Leaked",
>                    i, refcount1, refcount2);
> +
> +            if (num_fixed) {
> +                ret = update_refcount(bs, i << s->cluster_bits, 1,
> +                                      refcount2 - refcount1);

It would be nicer to use int64_t for i.  I haven't checked but it
makes me nervous to shift ints here.

Stefan