On Tue, 2018-09-25 at 13:15 +0800, Yi Min Zhao wrote: > > > > From the high level point of view, code outside of conf/ should, > > > > for the most part, not need to concern itself with zPCI at all: it > > > > would eg. ask for a PCI address to be allocated, and if the device > > > > in question can be a zPCI device then a zPCI extension address will > > > > be allocated for it as part of the same function call; the only > > > > part of qemu/ that should care about the zPCI address is the one > > > > generating the relevant command line arguments. > > > > > > > > Can you try and see whether this kind of API would work? > > > > > > I did a simple test. It worked. Do you prefer this way? > > > > Yes please, I'd very much like to see what that looks like and > > whether it addresses the problems caused by the ambiguity of the > > API we've used until now. > > So do you mean we use this kind of API in next version for review? Yes, please :) -- Andrea Bolognani / Red Hat / Virtualization -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
