Re: [PATCH] drm/virtio: Add window server support

From: Tomeu Vizoso <tomeu.vizoso@collabora.com>
To: Dave Airlie <airlied@gmail.com>
Cc: dri-devel <dri-devel@lists.freedesktop.org>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	David Airlie <airlied@linux.ie>, Jason Wang <jasowang@redhat.com>,
	LKML <linux-kernel@vger.kernel.org>,
	"open list:VIRTIO CORE,
	NET..."  <virtualization@lists.linux-foundation.org>,
	Gerd Hoffmann <kraxel@redhat.com>
Subject: Re: [PATCH] drm/virtio: Add window server support
Date: Fri, 12 Jan 2018 08:59:35 +0100	[thread overview]
Message-ID: <c25f1abf-d00d-c1ed-0aea-9a701c61a570@collabora.com> (raw)
In-Reply-To: <CAPM=9twT_jtfJteuUpRvu1Xn3nNfxG-5oLfcbHYVnHVeS9TzcA@mail.gmail.com>

On 01/12/2018 05:11 AM, Dave Airlie wrote:
>>
>> this work is based on the virtio_wl driver in the ChromeOS kernel by
>> Zach Reizner, currently at:
>>
>> https://chromium.googlesource.com/chromiumos/third_party/kernel/+/chromeos-4.4/drivers/virtio/virtio_wl.c
>>
>> There's two features missing in this patch when compared with virtio_wl:
>>
>> * Allow the guest access directly host memory, without having to resort
>> to TRANSFER_TO_HOST
>>
>> * Pass FDs from host to guest (Wayland specifies that the compositor
>> shares keyboard data with the guest via a shared buffer)
>>
>> I plan to work on this next, but I would like to get some comments on
>> the general approach so I can better choose which patch to follow.
> 
> Shouldn't qemu expose some kind of capability to enable this so we know to
> look for the extra vqs?

Sounds good. I'm unsure though on whether it should be done 
unconditionally if the hypervisor's code supports this, or if only if 
the user pass the -proxy-wayland switch and the hypervisor was able to 
open the socket to the compositor. I'm leaning towards the latter.

> What happens if you run this on plain qemu, does it fallback correctly?

Will work on this.

> Are there any scenarios where we don't want to expose this API because there
> is nothing to back it.

I'm not sure what the overhead of the extra queues is, but I guess the 
ioctls could immediately return -ENODEV if the hypervisor doesn't have 
that capability.

Happy to see that there aren't any major concerns with the general approach.

Thanks,

Tomeu