On 07-09-22, 19:02, Julien Grall wrote:
It is a very descriptive compatible :). And yes I realize this is the compatible chosen by upstream.
:)
So this is exposing a GPIO interrupt controller but IIUC we will also need to describe the devices in the device-tree. Can you outline how you would expect it to work?
The same is true for busses like I2C and SPI, which may have client devices in the DT as well.
Though I don't use them currently in my setup where I need to test the backends. I am able to test both I2C and GPIO dynamically from the guest, without any client devices in the DT.
To put some context, this is related to what I wrote in patch #1. How much code will be necessary in libxl to add new device?
I am not the maintainer of this code, but I would expect some example how this can be used once this merged. This would help us to understand whether the interface is correct.
I understand what you are looking for and that we may be required to pass the relation of a client device with the GPIO controller via the "virtio" entry you suggested for a common device.
I don't have an answer for that at the moment.