On 5/12/2026 8:51 PM, Mathieu Poirier wrote:
On Mon, 11 May 2026 at 12:18, Andrew Lunn <[email protected]> wrote:
Arnaud, Beleswar, Andrew and I are all advocating for one endpoint per
GPIO controller. The remaining issue it about the best way to work
out source and destination addresses between Linux and the remote
processor. I'm running out of time for today but I'll return to this
thread with a final analysis by the end of the week.
How many of the participants here will be in Minneapolis next week for
the Embedded Linux Conference? There is even a talk about this:
https://osselcna2026.sched.com/event/2JQpx/building-virtual-drivers-with-rpmsg-key-design-principles-challenges-trade-offs-beleswar-prasad-padhi-texas-instruments?iframe=yes&w=100%&sidebar=yes&bg=no
Maybe we can get together and decide on the final design after the
session.
I will not be in Minneapolis next week. At this point I think things
are converging into 2 main takeaways:
1) A serious refactoring of the protocol to include only what is
available in the virtio-gpio specification [1].
2) The specification of GPIO controller number in an extension of the
namespace announcement [2].
Fair enough. I am also aligned to use this solution with the support for
wildcard name service matching.
Thanks,
Beleswar
Shenwei proposed embedding the GPIO controller number in the
endpoint's source address [3], something I'm ambivalent about and
still have to look into. I also have to read Tanmay's latest
comments. I'm hoping to be done with all that by the end of the week.
With the above (1) and (2), a new patchset will be required to reset
this thread.
Thanks,
Mathieu
[1]. https://lwn.net/ml/all/afjyH5JT0JS2j0L5@p14s/
[2]. https://lwn.net/ml/all/afzIABSh1xtMEGbf%40p14s/
[3].
https://lwn.net/ml/all/paxpr04mb9185bfa6e7375fad0b15b02189...@paxpr04mb9185.eurprd04.prod.outlook.com/
Andrew
