On Mon, Jun 04, 2012 at 03:15:27PM -0400, Alan Stern wrote:
> On Mon, 4 Jun 2012, Felipe Balbi wrote:
>
> > On Mon, Jun 04, 2012 at 11:19:06AM -0400, Alan Stern wrote:
> > > On Mon, 4 Jun 2012, Felipe Balbi wrote:
> > >
> > > > This patch introduces support for test modes as
> > > > defined by the USB OTG & Embedded host Specification.
> > > >
> > > > Note that we _must_ be able to suspend & resume ports
> > > > as we wish, so this driver depends on USB_SUSPEND and
> > > > makes use (well, somewhat abuses) usb autopm features
> > > > to achieve what we want.
> > >
> > > Have you tested the autopm stuff? I don't think it will work the way
> > > you've got it.
> >
> > well, since I don't have the device myself, someone from a testing lab
> > has tested this driver. Any ideas how to implement that part with Linux?
> > That's a hard-requirement for USB OTG Certification, btw.
>
> Yes; it needs to be addressed.
>
> I believe it will be best to bypass the runtime PM framework. Either
> send the necessary requests directly to the parent hub, or let's add a
> test-mode API to the hub driver to do the same thing.
that would be much much nicer as it could also be extended later to
support USB[23]0CV test modes, so we could finaly run pre-certification
tests on Linux with an open source tool.
> In some ways this is simpler than an ordinary device suspend. For
> example, you're not concerned about remote wakeup, right?
that's correct.
> > > > +static void otg_eh_disconnect(struct usb_interface *intf)
> > > > +{
> > > > + struct usb_device *udev;
> > > > +
> > > > + udev = interface_to_usbdev(intf);
> > > > +
> > > > + usb_lock_device_for_reset(udev, intf);
> > >
> > > You don't need the locking around usb_reset_device(). Disconnect
> > > methods are called with the device lock already held.
> >
> > Will remove, but it's not what the comment on that function states. Or
> > at least it was a little dubious.
>
> The comment just above the function's kerneldoc has a slightly clearer
> explanation.
missed that one (sigh). Will read it tomorrow.
tks
--
balbi
Attachment:
signature.asc
Description: Digital signature
