On Fri, 29 Aug 2014, Peter Chen wrote: > > On Thu, 28 Aug 2014, Alan Stern wrote: > > > > > Okay, so we need to add a "vbus_is_on" flag to the usb_gadget > > > structure. The gadget driver will set this flag in its connect > > > callback and clear the flag in its disconnect callback. If the UDC > > > driver doesn't have a Vbus handler, it will invoke the connect > > > callback just once during startup, and it will invoke the disconnect > > > callback just once during removal. > > > > After more thought, if the vbus_is_on flag is going to be controlled by > > the gadget driver rather than the UDC driver, then it really belongs in > > the gadget driver's private data structure and not in struct usb_gadget. > > Also, a better name for the flag would be "is_connected". > > > > Yes, we should have some places to change "is_connected", that is one of > thing we are discussing. > > > > The gadget driver should include an adjust_pullup() routine. It > > > should get called whenever a function is activated or deactivated, and > > > also by the connect and disconnect callbacks. The logic is simple: > > > > > > If (any functions are deactivated) > > > turn off the pullup > > > else if (gadget->vbus_is_on) > > > turn on the pullup > > > else > > > turn off the pullup > > > > This logic can be compressed a little: > > > > if (!is_connected || any functions are deactivated) > > turn off the pullup > > else > > turn on the pullup > > > > Thanks, Alan. > > I have a similar implementation at below, > > http://www.spinics.net/lists/linux-usb/msg111976.html Yes, that seems to be moving in the right direction. You still need to change the other gadget drivers, and the API should be changed from composite_connect(gadget, flag) to composite_connect(gadget) and composite_disconnect(gadget). On Thu, 28 Aug 2014, Felipe Balbi wrote: > you shouldn't call usb_gadget_disconnect() from ->disconnect(). It's the > other way around. You get a disconnect IRQ, then you call gadget > driver's ->disconnect(), but gadget driver doesn't need to ask UDC to > remove pullup at that time. It's true that the gadget driver shouldn't have to ask for this, but doing so doesn't hurt. On the other hand, the gadget driver definitely _does_ have to tell the UDC driver to enable the pullup when a connect occurs. Otherwise the UDC driver won't know if the gadget's userspace components are ready. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
