On Mon, Sep 09, 2013 at 08:11:25PM +0200, Paolo Bonzini wrote: > Il 09/09/2013 20:06, Jan Kiszka ha scritto: > > archive, was in the context of lock-less MMIO dispatching), and the > > consensus back then was that device-to-device DMA is generally a bug > > that is not worth supporting in all its beauty. But if you know a > > concrete scenario / guest where it matters, that would bring in a new > > aspect. > > Well, one I know is > > 10 SCREEN 1 > 20 COLOR 1: CIRCLE (160, 100), 80 > 30 PRINT "Hello QEMU!" > 40 DEF SEG=&hB800 > 50 BSAVE "FOO.PIC", 16000 > RUN > > Not sure if that counts as something that matters, or even if it works > right now with virtio-blk. > > Paolo
I don't think it matters for virtio - it's a PV device and we can reasonably limit what works and what doesn't work there. It makes some sense for assigned devices (I think VFIO supports this to some extent) and we might at some point start emulating devices that do this. But mostly it's a driver bug, and all we need to do is make sure it doesn't cause security issues. -- MST
