On 06/01/2015 22:54, Alexander Graf wrote: > > > On 06.01.15 21:41, Paolo Bonzini wrote: >> >> >> On 06/01/2015 20:48, Paolo Bonzini wrote: >>> I like the way you structured the series! >>> >>> Reviewed-by: Paolo Bonzini <pbonz...@redhat.com> >> >> Hmm, actually doesn't this break -machine usb=no? > > I think it does, but I don't think we really need to care. We can just > assume that the 64bit mac99 always has USB and it's impossible to turn off. > > For sPAPR, we only ever enable USB as convenience option when VGA is > enabled. I think it's a pretty fair thing to do there and simply ignore > usb=no. If you really need it, just don't use -vga. > > At the end of the day, I don't think anyone will notice. I'd definitely > prefer to keep the code simple.
Yeah, I agree. If it's hard to fix, this patchset is okay. But if it's not too hard (and I don't think it is, especially if machine_parse is centralized as in my reply to "[PATCH] vl.c: fix regression when reading machine type from config file") the old semantics made more sense. Paolo > However, I think looking forward we'll want to spawn an XHCI adapter > rather than OHCI. It's just a lot cheaper to emulate. But that's out of > scope of this patch set.
