Il 03/06/2014 14:21, Sander Eikelenboom ha scritto:
The last part should perhaps be fixed in the driver/firmware (for instance the assumption the device is always device 00:02.0, that's valid for real intel hardware, but not necessary for passthrough or emulation). A requirement for using the latest firmware/drivers for passthrough to work could be acceptable i guess ?
The problem is worse than that. The assumptions are that the driver can make choices based on the device/vendor IDs of 00:1f.0, and that the driver can read/write to the config space of 00:00.0.
Paolo
