On 3/6/19 9:33 AM, Vladimir Sementsov-Ogievskiy wrote:
>> + /* Flush bitmaps */
>> + if (s->nb_bitmaps) {
>> + Error *local_err = NULL;
>> +
>> + /* Usually, bitmaps get resized after this call.
>> + * Force it earlier so we can make the metadata consistent. */
>> + bdrv_dirty_bitmap_truncate(bs, offset);
>> + qcow2_flush_persistent_dirty_bitmaps(bs, &local_err);
>> + if (local_err) {
>> + ret = -EINVAL;
>> + goto fail;
>> + }
>> + }
>
> Why to flush after resize? Bitmaps will be IN_USE in the image anyway...
>
> Could we implement resize without flush first, it would be one patch + test?
> And then consider
> flushing in separate?What happens with migration if we don't flush the new size to disk, so the on-disk format has a different size than the in-memory version? -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature
