Hi Linus,

On Wed, Oct 14, 2020 at 2:43 PM Rob Herring [off-list ref] wrote:

On Tue, Oct 13, 2020 at 8:27 AM Linus Walleij [off-list ref] wrote:

quoted

On Wed, Oct 7, 2020 at 9:58 PM Martin Blumenstingl
[off-list ref] wrote:

quoted

On Wed, Oct 7, 2020 at 11:19 AM Linus Walleij [off-list ref] wrote:

quoted

On Sun, Oct 4, 2020 at 8:00 PM Martin Blumenstingl
[off-list ref] wrote:

quoted

+properties:
+  compatible:
+    enum:
+      - pci1b6f,7023
+      - pci1b6f,7052

I think it is better to let the PCI driver for the whole hardware in
drivers/usb/host/xhci-pci.c probe from the PCI configuration space
numbers, and then add a gpio_chip to xhci-pci.c.

to have everything consistent I will move the binding to
Documentation/devicetree/bindings/usb

I do not understand why a PCI device would need a DT binding
at all. They just probe from the magic number in the PCI
config space, they spawn struct pci_dev PCI devices, not the
type of platform devices coming out of the DT parser code.
No DT compatible needed.

Same reason for all the discoverable buses need bindings. There can be
parts that are not discoverable or connections with non-discoverable
nodes. There's also cases where the discoverable device has to be
powered, reset deasserted, clocks enabled, etc. first to be
discovered.

If the GPIOs here had connections elsewhere in the DT, then we have to
describe the provider in DT.

this is exactly what I need it for: that platform has hand-written
.dts files and I need to wire up a GPIO LED


Best regards,
Martin