On Fri, 2013-12-13 at 11:10 +0800, Xuekun Hu wrote: > Alex, thanks. > > Yes, I'm talking about the pci device assignment. What about the vfio > performance, close to the pci pass-through performance? vfio is pci passthrough. It's a re-design of it. Performance is equivalent. > Currently I'm trying to understand what's the behind thinking about > the limited config space, security or sth else? Legacy kvm device assignment has traditionally restricted access to device config space. I suspect it's a matter of necessary functionality and vague ideas that it provides greater security. Device assignment would be much easier to implement if devices strictly followed the spec, but we know that's not the case. Thanks, Alex > On Fri, Dec 13, 2013 at 10:55 AM, Alex Williamson > <alex.williamson@xxxxxxxxxx> wrote: > > On Fri, 2013-12-13 at 10:34 +0800, Xuekun Hu wrote: > >> Hi, All > >> > >> I found qemu/kvm has limited pci config space exposed to guest device > >> driver to access. I’m wondering is there any reason why not expose the > >> whole pci config space to guest, since some kind of devices has > >> special register in that area. > > > > If you're talking about device assignment, vfio allows access to all of > > config space. Legacy KVM device assignment is deprecated. Thanks, > > > > Alex > > -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
