Hi,

quoted

qemu can use memfd to allocate guest ram.  Now, with the help of
udmabuf, qemu can create a *host* dma-buf for the *guest* graphics
buffer.

Guess each physical address in the iovec in
VIRTIO_GPU_CMD_RESOURCE_ATTACH_BACKING can be passed as the offset in the
udmabuf_create_item struct?

Exactly.

https://git.kraxel.org/cgit/qemu/commit/?h=sirius/udmabuf&id=515a5b9f1215ea668a992e39d66993a17a940801

Are you thinking of anything else besides passing the winsrv protocol across
the guest/host boundary? Just wondering if I'm missing something.

The patch above uses the dmabuf internally in qemu.  It simply mmaps it,
so qemu has a linear representation of the resource and can use it as
pixman image backing storage without copying the pixel data.

So it is useful even without actually exporting the dmabuf to other
processes.

cheers,
  Gerd

PS: Any chance you can review the v7 patch?