Here is a patch that implements the node device destroy API call, as well as the suggestions that Dan made in his feedback on the patch implementing create. The patch also makes it not necessary for the caller of the node device create function to specify a name for the device and a host number for the adapter in the XML. The change to do so is kind of kludgy, though, IMO, as it involved changing the function signature of chain of internal function calls to differentiate a device that is being created from an existing device. I'm torn as to whether I think that's a better solution than requiring the caller of the nodedev create API to specify a name that will be ignored. One benefit of requiring the user to specify the name is that it is used in some error messages, so I can argue it both ways. Opinions very welcome. I'd particularly like someone in the Solaris world to make sure that the code builds as there is some #ifdef __linux__ in the patch. I tried to spin up Opensolaris to try it out but ran up against bug 6784591 and a lack of time. I've also tried to make it relatively easy for someone to add support for non-Linux OSes, so a patch there would be appreciated. Dave -- Libvir-list mailing list Libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list