Re: [PATCH 0/4] Add SCMI vhost-user VIRTIO device

[Vincent Guittot]

On Wed, 31 May 2023 at 13:57, Milan Zamazal <mzamazal@redhat.com> wrote:
>
> Alex Bennée <alex.bennee@linaro.org> writes:
>
> > Milan Zamazal <mzamazal@redhat.com> writes:
> >
> >> This patch series adds a vhost-user VIRTIO device for SCMI.
> >> It's similar to other similar vhost-user VIRTIO devices.
> >>
> >> I'm aware of the work in progress by Alex Bennée to simplify similar 
> >> devices
> >> and avoid excessive code duplication.  I think the SCMI device support
> >> doesn't bring anything special and it can be rebased on the given work
> >> if/once it is merged.
> >
> > \o/ - I'll try and get the next iteration done in the next few weeks.
>
> Great, looking forward.
>
> > Out of interest have you tested your scmi backend with those patches? Oh
> > and also which backend are you using?
>
> I have tested the patches with my proof-of-concept SCMI backend
> (emulating a fake sensor) based on rust-vmm and a Linux guest OS.  I
> plan to change the prototype into something publishable in the next
> weeks and to post patches to rust-vmm/vhost-device.
>
> > Vincent did a bunch of work over the last year or so on SCMI although in
> > his case the backend was a RTOS running in a separate domain using some
> > shared memory between the SCMI domain and a couple of guests.
>
> I have seen recordings of some related presentations.  The work is
> primarily based on crosvm, right?  I'd be interested in what kind of

I have used qemu for my tests

> SCMI interface the domain provides and whether it would make sense to be
> able to connect it with the QEMU SCMI device via some vhost-user daemon,
> an already existing one, if any, and/or the one I work on.

My PoC is based on the SCP-firmware that is usually running on a
dedicated power coprocessor. We have extended it to run in several
env. The PoC runs the SCP firmware in a VM and uses a vhost user scmi
bridge to connect the guest vm with this virtual power coprocessor.
One reason for using SCP-firmware is that It follows SCMI specs and
its updates so I don't have to re-implement it myself and can leverage
on their devs

For test purpose I also have enabled it as a simple linux vhost user daemon

>
> >> Milan Zamazal (4):
> >>   hw/virtio: Add boilerplate for vhost-user-scmi device
> >>   hw/virtio: Add vhost-user-scmi-pci boilerplate
> >>   tests/qtest: Fix a comment typo in vhost-user-test.c
> >>   tests/qtest: enable tests for virtio-scmi
> >>
> >>  MAINTAINERS                         |   7 +
> >>  hw/virtio/Kconfig                   |   5 +
> >>  hw/virtio/meson.build               |   2 +
> >>  hw/virtio/vhost-user-scmi-pci.c     |  68 +++++++
> >>  hw/virtio/vhost-user-scmi.c         | 306 ++++++++++++++++++++++++++++
> >>  include/hw/virtio/vhost-user-scmi.h |  30 +++
> >>  tests/qtest/libqos/meson.build      |   1 +
> >>  tests/qtest/libqos/virtio-scmi.c    | 174 ++++++++++++++++
> >>  tests/qtest/libqos/virtio-scmi.h    |  34 ++++
> >>  tests/qtest/vhost-user-test.c       |  46 ++++-
> >>  10 files changed, 672 insertions(+), 1 deletion(-)
> >>  create mode 100644 hw/virtio/vhost-user-scmi-pci.c
> >>  create mode 100644 hw/virtio/vhost-user-scmi.c
> >>  create mode 100644 include/hw/virtio/vhost-user-scmi.h
> >>  create mode 100644 tests/qtest/libqos/virtio-scmi.c
> >>  create mode 100644 tests/qtest/libqos/virtio-scmi.h
>