Re: [PATCH 2/3] usb: type-c: USB Type-C Connector System Software Interface

[Greg KH]

On Wed, Feb 10, 2016 at 12:30:42PM +0200, Heikki Krogerus wrote:
> On Tue, Feb 09, 2016 at 10:21:55AM -0800, Greg KH wrote:
> > On Tue, Feb 09, 2016 at 07:01:22PM +0200, Heikki Krogerus wrote:
> > > USB Type-C Connector System Software Interface (UCSI) is a
> > > specification that defines registers and data structures
> > > used to interface with the USB Type-C connectors on a system.
> > > 
> > > The specification is public and available at:
> > > http://www.intel.com/content/www/us/en/io/universal-serial-bus/usb-type-c-ucsi-spec.html
> > > 
> > 
> > What does this driver / code actually do?  Why is it needed?  What
> > interface to the rest of the kernel / userspace does it provide?
> 
> I will fix this describe these things in the commit message. I'll just
> but some UCSI background in case somebody is interested. So UCSI is in
> practice a standard for USB Type-C controllers..
> 
> UCSI is the control interface for USB Type-C connectors (regardless
> was USB PD supported or not) in MS Windows, so most likely all new HW
> platforms designed to work also with Windows that are equipped with
> USB Type-C will have UCSI device for controlling the USB Type-C ports.

There's many millions of devices with type-C without Windows on them, so
don't count on this being everywhere :)

> In some cases the hardware for Type-C will be just a PHY like fusb30x
> on these platforms (it's cheaper then USB PD or complete USB Type-C
> controller), but in those cases the PHY is probable attached to an EC
> or is completely controlled by system FW like BIOS together with any
> USB PD communication in cases where USB PD is supported, and is in any
> case not visible to the OS. Instead UCSI device is exposed to the OS
> to give it means to apply its policies to the USB Type-C port.
> 
> > Why would we care about this?
> 
> I'll try to explain why it's important to export the control of USB
> Type-C ports to the user space in my answer to your comments to the
> first patch of this series, the one introducing the class.
> 
> But surely everybody agrees that decision about the policies regarding
> USB Type-C ports, like which data role to use, do we charge or are we
> letting the other end charge, etc., belongs to the user?

No, I don't agree.  It's still unknown if userspace can react fast
though to these types of "policy" changes.  I've heard from some
manufacturers that the response time needed is something that we can't
leave to userspace.

And along those lines, do you have a working userspace user of this
interface?  We don't create interfaces without a user, especially given
that it takes a long time to ensure that a user/kernel api actually is
correct.  We would need to see that to ensure that this kernel
implementation is "correct" and working properly.

> If you plug
> your phone to your desktop, I would imagine that you want to see the
> phone primarily as the USB device and the desktop as host, and to
> achieve the device role, you don't want to be forced to unplug/replug
> your phone from the desktop until you achieve device role, right?

Why is unplugging somehow required?

thanks,

greg k-h