On Thu, 02 Aug 2012 15:46:48 +0200
Paolo Bonzini <pbonz...@redhat.com> wrote:
> Il 02/08/2012 15:27, Markus Armbruster ha scritto:
> >> > + bdev_list = qmp_query_block(NULL);
> >> > + for (bdev = bdev_list; bdev; bdev = bdev->next) {
> >> > + if (!strcmp(bdev->value->device, device) &&
> >> > + blockinfo_is_encrypted(bdev->value) &&
> >> > + !blockinfo_key_is_set(bdev->value)) {
> >> > + hmp_change_ask_user_key(mon, bdev->value);
> >> > + break;
> >> > + }
> >> > + }
>
> Is this anything that an external application can reproduce?
Yes, that should be possible.
But thinking a bit more about this, the real question is whether we want
them to do it. I guess not, as the fact that qmp_bdrv_open_encrypted()
doesn't close the bs on error is probably a bug.
> Perhaps we need to keep QERR_DEVICE_ENCRYPTED even if libvirt does not
> use it, or at least provide an alternative mechanism (e.g. an event) to
> realize its effect.
Yes, maybe that's better. hmp_cont() wouldn't change much though, as
it still needs to figure out which device needs a key.
