Messages in this thread | | | Subject | Re: [PATCH 1/5] dt-bindings: connector: add optional properties for Type-B | From | Hans de Goede <> | Date | Mon, 11 Mar 2019 12:00:10 +0100 |
| |
Hi,
On 11-03-19 09:06, Hans de Goede wrote: > Hi, > > On 11-03-19 06:33, Chunfeng Yun wrote: >> Hi, >> >> On Fri, 2019-03-08 at 13:07 +0100, Hans de Goede wrote: >>> Hi, >>> >>> On 08-03-19 07:13, Chunfeng Yun wrote: >>>> Add id-gpios, vbus-gpios, vbus-supply and pinctrl properties for >>>> usb-b-connector >>>> >>>> Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> >>>> --- >>>> .../devicetree/bindings/connector/usb-connector.txt | 10 ++++++++++ >>>> 1 file changed, 10 insertions(+) >>>> >>>> diff --git a/Documentation/devicetree/bindings/connector/usb-connector.txt b/Documentation/devicetree/bindings/connector/usb-connector.txt >>>> index a9a2f2fc44f2..7a07b0f4f973 100644 >>>> --- a/Documentation/devicetree/bindings/connector/usb-connector.txt >>>> +++ b/Documentation/devicetree/bindings/connector/usb-connector.txt >>>> @@ -17,6 +17,16 @@ Optional properties: >>>> - self-powered: Set this property if the usb device that has its own power >>>> source. >>>> +Optional properties for usb-b-connector: >>>> +- id-gpios: gpio for USB ID pin. >>> >>> What about boards where the ID pin is *not* connected to a GPIO, >>> but e.g. to a special pin on the PMIC which can also detect >>> an ACA adapter ? Currently this case is handled by extcon >>> drivers, but we have no way to set e.g. vbus-supply for the >>> connector. Maybe in this case the usb-connector node should >>> be a child of the PMIC node ? >> Yes, it would be, PMIC is in charger of detecting the status of ID pin > > Ok, then I think this should be documented too. > >>> And in many cases there also is a mux to switch the datalines >>> between the host and device(gadget) controllers, how should >>> that be described in this model? See the new usb-role-switch >>> code under drivers/usb/roles >>> >>> In some cases the mux is controlled through a gpio, so we >>> may want to add a "mux-gpios" here in which case we also >>> need to define what 0/1 means. >> I'm not sure, the mux seems not belong to this connector, >> and may need another driver to register usb-role-switch, >> similar to: >> >> [v2,2/2] usb: typec: add typec switch via GPIO control >> https://patchwork.kernel.org/patch/10834327/ > > Right the mux/role-switch will need a driver, but the "owner" > of the usb_connector, e.g. the PMIC or the owner of the > id GPIO pin needs to know which device is the role-switch so > that it can set the role correctly based on the id-pin. > > Your binding already contains Vbus info, allowing the owner > of the usb_connector to enable/disable Vbus based on the id-pin, > but the owner will also be responsible for setting the role-switch. > > Note we cannot simply assume there will be only one role-switch, > we really need some link from the usb_connector to the role-switch > (or if it is a GPIO driven role-switch simply a role-switch-gpios > member in the usb_connector).
I see now in your "[PATCH v3 1/2] dt-bindings: usb: add documentation for typec switch via GPIO" that you plan to use a Documentation/devicetree/bindings/graph.txt style binding for this, adding a remote-endpoint to the orientation-switch (in that patch), or to the role-switch.
But a Type-C port can have both an orientation-switch and a role-switch (and a mux if it supports e.g. DP-altmode), how is the driver driving the device which has the actual usb_c_connecter child-node to which the remote-endpoints for both the orientation- and the role-switch points supposed to figure out which is which ?
Also it feels the wrong-way around to me to have the orientation-switch point to the usb_c_connector, to me it would make more sense to have the usb_c_connector point to a port on the orientation-switch.
Regards,
Hans
| |