On Thu, Apr 26, 2018 at 11:00:28AM -0600, Alex Williamson wrote: > On Thu, 26 Apr 2018 18:40:40 +0300 > "Michael S. Tsirkin" <m...@redhat.com> wrote: > > > On Thu, Apr 26, 2018 at 05:31:48PM +0200, Gerd Hoffmann wrote: > > > On Thu, Apr 26, 2018 at 06:10:48PM +0300, Michael S. Tsirkin wrote: > > > > On Tue, Apr 24, 2018 at 08:38:57AM +0200, Gerd Hoffmann wrote: > > > > > mdpy is a sample pci device for vfio-mdev. Not (yet) merged upstream, > > > > > patch available here: > > > > > > > > > > https://www.kraxel.org/cgit/linux/commit/?h=vfio-sample-display&id=6fd86cff3d7df38ab89625b16fdd6434b1c18749 > > > > > > > > > > Cc: Alex Williamson <alex.william...@redhat.com> > > > > > Signed-off-by: Gerd Hoffmann <kra...@redhat.com> > > > > > > > > Reviewed-by: Michael S. Tsirkin <m...@redhat.com> > > > > > > > > I presume this should be merged together with the > > > > device implementation in qemu, or do you want this > > > > in tree sooner? > > > > > > There isn't a device implementation in qemu. It's a mdev device > > > provided by the kernel. Can be used with qemu using vfio. > > > > Why do we need a define in pci.h then? > > Because we want to allocate the device ID rather than squat on an > unused one and risk that it later conflicts with an another device. Do > we track device ID allocations from the RH vendor ID other than here in > QEMU? Thanks, > > Alex
We have pci-ids.txt, yes. Unused macros just might get pruned down the line. If you like, add a comment in pci.h saying 0xf is reserved for mdpy. -- MST
