Quoting Peter Chen (2016-09-26 21:53:58)

On Mon, Sep 26, 2016 at 11:44:50AM -0700, Stephen Boyd wrote:

quoted

Quoting Peter Chen (2016-09-25 20:29:27)

quoted

On Thu, Sep 22, 2016 at 11:51:02AM -0700, Stephen Boyd wrote:

quoted

Quoting Peter Chen (2016-09-16 18:16:05)

quoted

On Wed, Sep 14, 2016 at 01:55:02AM -0700, Stephen Boyd wrote:

quoted

Quoting Stephen Boyd (2016-09-13 18:42:46)

      soc {
              usb at 78d9000 {
                      extcon = <&usb_id>, <&usb_id>;

Why you have two same extcon phandler? From my mind, one should id,
another should is vbus. Besides, I find extcon-usb-gpio.c is lack of
vbus support, how you support vbus detection for
connection/disconnection with PC for your chipidea msm patch set?

This was already in the dts files for db410c. In the chipidea binding
one is for EXTCON_USB (vbus) and one is for EXTCON_USB_HOST (id). My
understanding is that extcon-usb-gpio.c sends events for both EXTCON_USB
and EXTCON_USB_HOST when the gpio changes state. vbus detection is not
that great on this board because we only have on gpio for this.

I think extcon-usb-gpio.c needs to extend for supporting vbus event,
otherwise, the micro-b cable's connect/disconnect will introduce
EXTCON_USB_HOST event, if you use two <&usb_idx> for both id and
vbus event.

Sorry, I'm lost now. extcon-usb-gpio.c already supports EXTCON_USB as an
event. Is the problem that we're using two of the same phandles in the
binding?

No, ID and VBUS are different events.

http://www.spinics.net/lists/linux-usb/msg147004.html

Agreed. But we only register for EXTCON_USB or EXTCON_USB_HOST for each
phandle in the list respectively based on the index of the phandle. So
with or without the patch you mention I don't see how it matters.

We do it this way to trigger the role switch when the cable is
disconnected. I think the problem with just having the id phandle is
that we'll never see the vbus event, and so the role switch doesn't
work.