On Tue, Dec 20, 2011 at 03:25:56PM +0200, Heikki Krogerus wrote:
> On Tue, Dec 13, 2011 at 01:21:17PM +0200, Felipe Balbi wrote:
> > > Why there needs struct usb_otg at struct usb_phy?
> > > From my point, at general otg driver, it doesn't need to touch phy.
> > > The phy are only needed to operated at vendor's otg driver.
> > > So, I don't know why below is not ok:
> > >
> > > > struct twl6030_usb {
> > > > > > > + struct usb_phy phy;
> > > > > > > + struct usb_otg otg;
> > >
> > > struct usb_otg {
> > > unsigned otg_specific;
> > >
> > > >
> > > struct otg_phy {
> > > unsigned phy_specific;
> > > }
> > >
> > > Besides, for some SoCs, there is no otg driver at mainline until now, but
> > > they need to get usb_phy.
> >
> > the thing is that OTG adds a few extra bits and pieces to the PHY. For
> > example we have extra states on the USB state machine, we have a few
> > analog comparators (for VBUS and ID pin) and so on. So, when we _do_
> > have OTG, some of it is handled by the PHY.
>
> One more question before I make the v9. If you have two PHYs but only
> one of the is OTG, how do you make sure the other PHY driver can not
> take advantage of OTG utility?
We have been thinking about adding a phy_type kinda thing, or maybe a
"features" flag ?!?
--
balbi
Attachment:
signature.asc
Description: Digital signature
