[Qemu-devel] Re: [PATCH 0/4] Block DMA helpers (v2)

[Anthony Liguori <aliguori@us.ibm.com>]

Avi Kivity wrote:
> We could replace
>
>    struct iovec in_sg[VIRTQUEUE_MAX_SIZE];
>    struct iovec out_sg[VIRTQUEUE_MAX_SIZE];
>
> by QEMUSGList, which would be a minor improvement, but I thought 
> things were very different for some reason.

I'm not sure that it's just a minor improvement.  To maintain this 
interface, we have to map the full request.  If we exposed QEMUSGLists, 
then each device could choose the mapping policy.  Since there's no good 
reason to use virtio to DMA MMIO memory, I'm unsure which is the best 
approach.

> An example would be how the code is structured now (as opposed to how 
> I thought it was structured).

Okay, I was certain I was missing something obvious :-)

Regards,

Anthony Liguori