On 4/23/25 09:28, Zhenzhong Duan wrote:
Hi,
This series addresses Cédric's suggestion[1] and Donald's suggestion[2] to
move realize() call after attach_device().
This way avoid the need to introduce realize_late() to further complex the
interface in nesting series.
[1] https://lists.gnu.org/archive/html/qemu-devel/2025-04/msg01211.html
[2] https://lists.gnu.org/archive/html/qemu-devel/2025-04/msg00898.html
Test:
net card passthrough and ping test
hotplug/unplug
Based on vfio-next(856f36c005).
Thanks
Zhenzhong
Changelog:
v2:
- drop the idea to save host iommu capabilities in VFIODevice.caps
- introduce a new function to create and realize hiod
- remove hiod_typename property
Zhenzhong Duan (5):
vfio/iommufd: Make a separate call to get IOMMU capabilities
vfio/iommufd: Move realize() after attachment
vfio/container: Move realize() after attachment
vfio: Cleanup host IOMMU device creation
vfio: Remove hiod_typename property
include/hw/vfio/vfio-container-base.h | 3 ---
include/hw/vfio/vfio-device.h | 3 ++-
hw/vfio/container.c | 25 ++++++++++++--------
hw/vfio/device.c | 33 ++++++++++++---------------
hw/vfio/iommufd.c | 31 +++++++++++++------------
5 files changed, 47 insertions(+), 48 deletions(-)
Reviewed-by: Cédric Le Goater <c...@redhat.com>
Thanks,
C.
