Re: [PATCH V3 1/3] gpio: Add virtio-gpio driver

[Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>]

On Wed 16 Jun 10:52 CDT 2021, Enrico Weigelt, metux IT consult wrote:

> On 16.06.21 05:30, Bjorn Andersson wrote:
> 
> > Combined with the virtio-i2c effort this could provide an alternative by
> > simply tunneling the busses and GPIOs into Linux and use standard iio
> > drivers, for cases where this suits your product requirements better.
> 
> So, you wanna use virtio as logical interface between the two CPUs ?
> Interesting idea. Usually folks use rpmsg for those things.
> 

rpmsg is a layer on top of virtio, so this would be an extension of the
existing model.

There's been discussions (and I believe some implementations) related to
bridging I2C requests over rpmsg, but I think it's preferable to
standardize around the virtio based bearer directly.

> What is running on the secondary CPU ? Some OS like Linux or some bare
> metal stuff ? What kind of CPU is that anyways ?
> 

These ideas revolves around platforms that implements something like the
"Android Sensor Hub", which provides some resource constraint
co-processor that deals with sensor device interaction and processing of
the data without waking up the power-hungry ARM cores.

Given the focus on power consumption I would guess that these are not
going to run Linux. Core-wise I've seen this implemented using primarily
ARM and Hexagon cores.

Regards,
Bjorn