On 09/08/2015 11:52 PM, Eric Blake wrote: > On 09/08/2015 03:10 AM, Wen Congyang wrote: > >>> Design-wise, I think we really want to have the add-child operation be >>> handed a pre-opened BDS, rather than the options dictionary to open the >>> BDS itself. That is, we should use the existing blockdev-add (and >>> enhance it to support everything) to open the BDS, and then this command >>> should just attach that BDS as the new child (which is why it IS >>> important that we validate that the new BDS being added doesn't create >>> an invalid loop). >>> >> >> How to check it? The parent BDS can get all children. But the child doesn't >> know if it is some BDS's child. > > If I'm not mistaken, a child DOES know what its parent(s) are, once we > have Max's series for NULL BDS representing a BB without media. >
Which patch? I don't find it. Thanks Wen Congyang