IMO one of the biggest parts of infrastructure QEMU is missing for embedded stuff is a common (qdev) way of modeling DMA backchannels. Or devide to device channels or whatever you want to call it.
Would be nice if someone could bring this to qemu. Cheers
