Virtio Xen (WAS: Re: [Linux] [ARM] Granting memory obtained from the DMA API)

[Julien Grall <julien@xxxxxxx>]

On 20/08/2020 12:57, Simon Leiner wrote:
> Hi Julien,
Hi Simon,

> On 20.08.20 13:17, Julien Grall wrote:
> > There is a lot of interest to get Virtio working on Xen at the moment.
> > Is this going to be a new transport layer for Virtio?
>
> It is designed that way, yes. The current implementation (based on
> virtio_mmio.c) has a few limitations:
>   - Only the host side is implemented for Linux (We are currently only
> using bare metal domains for the device side - so the device
> implementation is based on OpenAMP[1])
>
> - It lacks some features, e.g. there is currently no device
> configuration space
>
> - It is tested only very narrowly (only for my use case which is RPMsg
> via the rpmsg_char kernel driver)
>
> As this was really just a byproduct of my main research topic, I'm
> currently not in touch with the virtio standards committee. But I'm
> happy to contribute my work if there is interest :-)
I have CCed a few folks that are interested in virtio. May I ask why did 
you create a new transport rather than using the existing one?
So far, there is an RFC to implement virtio-mmio for Xen on Arm (see 
[2]). But the idea of a Xen specific transport is discussed on the 
mailing list time to time. It would be more secure than existing 
transport, but I am worried about the adoption of the transport.
A new transport requires to modify all the OSes so they can work on Xen. 
This is fine for open-source OS, but it may be more difficult to get 
proprietary OS to invest in the new transport.
Do see any use of this transport outside of Xen?

Cheers,

[2] 
https://lore.kernel.org/xen-devel/1596478888-23030-1-git-send-email-olekstysh@xxxxxxxxx/
--
Julien Grall