On 08/01/2011 10:44 AM, Stefan Hajnoczi wrote:
On Mon, Aug 1, 2011 at 4:44 PM, Kevin Wolf<kw...@redhat.com> wrote:
Am 01.08.2011 17:28, schrieb Anthony Liguori:
On 08/01/2011 10:22 AM, Stefan Hajnoczi wrote:
I like the block_set approach.
Anthony, Kevin, Supriya: Any thoughts?
For the sake of overall QMP sanity, I think block_set_hostcache is
really our only option.
Ideally we should have blockdev_add, and blockdev_set would just take
the same arguments and update the given driver.
But we don't have blockdev_add today, so whatever works for your as a
temporary solution...
Anthony's point is that blockdev_set does not fit with QMP command
discoverability.
It doesn't, but if we had a 'plug_set' that worked for devices and any
type of backends, we could have a single introspection mechanism.
But we should really try to avoid having every backend implement their
own ways of doing things.
Regards,
Anthony Liguori
Stefan