Le mar. 1 sept. 2020 à 22:41, Loic PALLARDY <loic.pallardy@st.com> a écrit :




> -----Original Message-----

> From: boot-architecture <boot-architecture-bounces@lists.linaro.org> On

> Behalf Of Stefano Stabellini

> Sent: mardi 1 septembre 2020 22:19

> To: François Ozog <francois.ozog@linaro.org>

> Cc: boot-architecture@lists.linaro.org; system-dt@lists.openampproject.org;

> Tomas Evensen <tomase@xilinx.com>

> Subject: Re: [System-dt] Notes (brief) from System DT call: OpenAMP

> bindings

>

> On Tue, 1 Sep 2020, François Ozog via System-dt wrote:

> > On Tue, 1 Sep 2020 at 14:49, Tomas Evensen <tomase@xilinx.com> wrote:

> >       Here are some brief notes from the meeting.

> >

> >       If nothing else as a showcase on how much we miss Nathalie when she

> is on vacation 😉

> >

> >       Krzysztof Kepa – GE

> >       Mathieu Poirier – Linaro

> >       Etsam Anjum – Mentor

> >       Anrnoud Pouliquen – ST

> >       Loic Pallardy - ST

> >       Suman Anna - TI

> >       Dan Milea – Wind River

> >       Mark Dapoz – Wind River

> >       Tomas Evensen – Xilinx

> >       Stefano Stabellini – Xilinx

> >       Bruce Ashfield – Xilinx

> >       Ed Mooring – Xilinx

> >       Ben Levinsky - Xilinx

> >

> >       Stefano went through slides.

> >       Overall idea:

> >

> >         1.  Specify remoteproc channels with minimal information in System-DT

> in a way that is as common as possible for all vendors.

> >       In particular, we are avoiding to have to specify the same memory

> regions in more than one place. A resource group is specially

> >       marked to contain most information.

> >         2.  Use Lopper to create the vendor specific remoteproc specification

> >       Generating reserved-memory, etc. information that has duplicate

> information.

> >       For the time being, no plans to unify the vendor specific information

> that goes into the traditional device tree for Linux

> >         3.  Later (or in parallel), but without making it a gating item, start

> discussions on what we can do to unify the vendor

> >       specific information in the traditional DT

> >

> >       Discussion about how TI and ST might generate their specific information

> from this specification.

> >

> >       Stefano to send out examples and slides.

> >       Ben: Xilinx backend to Lopper is upstreamed and available as an

> example. Might change as upstreaming continues.

> >

> >       Question:

> >       * How do we configure VirtIO?

> >

> > Could you describe more the VirtIO question so that relationship with

> Linaro project Stratos is assessed?

>

> I think this point refers to various vdev<x> buffers described under

> /reserved-memory which are linked from RemoteProc. From the Xilinx

> bindings:

>

>     reserved-memory {

>               #address-cells = <1>;

>               #size-cells = <1>;

>               ranges;

>

>               rpu0vdev0vring0: rpu0vdev0vring0@3ed40000 {

>                       compatible = "xilinx,openamp-ipc-1.0";

>                       no-map;

>                       reg = <0x3ed40000 0x4000>;

>               };

>               rpu0vdev0vring1: rpu0vdev0vring1@3ed44000 {

>                       compatible = "xilinx,openamp-ipc-1.0";

>                       no-map;

>                       reg = <0x3ed44000 0x4000>;

>               };

>               rpu0vdev0buffer: rpu0vdev0buffer@3ed48000 {

>                       compatible = "xilinx,openamp-ipc-1.0";

>                       no-map;

>                       reg = <0x3ed48000 0x100000>;

>               };

>               rproc_0_reserved: rproc@3ed000000 {

>                       compatible = "xilinx,openamp-ipc-1.0";

>                       no-map;

>                       reg = <0x3ed00000 0x40000>;

>               };

>       };

ST proposed an RFC to decouple virtio rpmsg device from rproc one. In this proposal, a DT node is introduced to described HW associated to virtio device (share memory, mboxes...) in order to make description generic between different SoCs.
Is there a possible relationship with FF-A memory allocations ? I feel that memory reservations are better off dealt with through that mechanism.

As Stephano is proposing a generic way to populate DT kernel rproc node from system DT, I propose to think about virtio rpmsg device node definition (and why not any virtio device) in a second step.

Regards,

Loic

> _______________________________________________

> boot-architecture mailing list

> boot-architecture@lists.linaro.org

> https://lists.linaro.org/mailman/listinfo/boot-architecture

--
François-Frédéric Ozog | Director Linaro Edge & Fog Computing Group
T: +33.67221.6485
francois.ozog@linaro.org | Skype: ffozog