[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 0/3] input: linux evdev support
From: |
Paolo Bonzini |
Subject: |
Re: [Qemu-devel] [PATCH 0/3] input: linux evdev support |
Date: |
Wed, 9 Mar 2016 22:23:34 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.6.0 |
On 04/03/2016 11:25, Gerd Hoffmann wrote:
> Hi,
>
> This patch series adds support for reading input events directly from
> linux input devices instead of getting them from the UI (gtk/spice/...).
> This is useful if you pci passthrough your vga, because you don't need
> some otherwise dummy UI just to feed input into your guest. Chech the
> patch #1 commit message for all the details.
>
> It's been a while I posted the patches last time. Undusted them.
> Rebased to master. Adapted to some QAPI changes. Squashed in some
> bugfixes accumulated over time. Applied some testing using my new
> intel test box.
This is nice! I have used virtio-input-host to do some pseudo-multiseat,
but it was only for Linux guests.
However, instead of adding a new -input-linux option, could you make it
a QOM object which implements UserCreatable? Then you can add it with
something like "-object input-linux,path=/dev/input/input10" (perhaps
"input-evdev" would be more specific). This has three advantages:
1) you get hotplug for free;
2) you don't add yet another option to vl.c (btw patch 2 and 3 are not
updating the docs);
3) it's easier to add more backends, though the only ones that come to
mind are rather silly (e.g. input-msmouse could take a chardev and parse
the serial mouse protocol).
If you cannot use QOM, even just using "-inputdev
[backend=]linux,path=/dev/input/input10" would provide (3), but QOM
seems superior at the cost of a little more boilerplate in ui/input-linux.c.
Thanks,
Paolo