Gregory Haskins wrote: > On Sat, 2007-06-23 at 10:53 +0300, Avi Kivity wrote: > >> Gregory Haskins wrote: >> >>> In fact, my own implementation puts the ISA-userint model as hanging off >>> the kvm structure as well (which would be replaced with a similar PIC >>> model for layer-2 mode in the future). >>> >>> >>> >> The PIC is certainly a system resource, but at least in qemu it's >> connected to cpu 0 only. We want to preserve that. >> >> > > I think there might be some confusion here ;) The *outputs* of the PIC > in both QEMU and in-kernel-APIC are certainly wired to cpu0 (per the > model that qemu/bochs support). But the *inputs* are system-wide. In > qemu, you can call the global function "pic_set_irq" (which you could > construe as per-vm, since there is a 1:1 with qemu-process/vm). > Likewise, the inputs in both in-kernel implementation is also per vm by > hanging off of the kvm struct, as opposed to the vcpu struct. I hope > this helps to clarify. > >
This is my understanding too. I didn't suggest the PIC ioctls should be per-vcpu, just that current semantics should be preserved. ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel