Re: [RFC/PATCH 2/2] usb: Adding SuperSpeed support to dummy_hcd

[Sarah Sharp <sarah.a.sharp@linux.intel.com>]

On Tue, Oct 12, 2010 at 09:55:29AM -0400, Alan Stern wrote:
> On Tue, 12 Oct 2010, Brokhman Tatyana wrote:
> 
> > Hi Alan
> > 
> > Thanks you for your comments. Please see my reply inline.
> 
> > > In addition, I suspect the dummy_hcd driver structure shouldn't contain
> > > an address_device entry.  It should be present only in dummy_ss_hcd.
> > 
> > I'm not sure I follow...
> > According to the code and comments in hub.c address_device cb is used if
> > the host controller wishes to choose the device address itself instead of
> > the address chosen by the core.
> 
> Correct.
> 
> >  It seems to me that there is nothing
> > preventing the dummy_hcd from supplying such cb as well. Is there?
> > Having both HS and SS dummy_hcd determine the device address seems more
> > convenient for testing proposes.
> 
> For testing purposes, it is best to imitate the behavior of a real 
> device as closely as possible.  USB-2.0 host controllers do not assign 
> their own addresses to devices; they use the addresses provided by 
> usbcore.  I think this is also true for the low/full/high-speed 
> components of a USB-3.0 controller.

There is only one controller, the xHCI controller, that handles all
device speeds.  There are no companion controllers for a USB 3.0 host
controller.  The xHCI host controller uses the hardware assigned address
for all devices (LS/FS/HS/SS), not the usbcore address.

But I'll admit I don't understand what dummy-hcd is supposed to do, so
I'm a bit confused as to why it needs an address_device function.  Is it
supposed to be the gadget-side interface to the hardware that's
transmitting over USB (i.e. responding to the USB host)?  Or does it
have something to do with a gadget switching over to host mode for OTG?

Sarah Sharp