Re: [PATCH v5 4/6] usb: roles: add API to get usb_role_switch by node

[Heikki Krogerus]

On Mon, May 20, 2019 at 10:39:11AM +0800, Chunfeng Yun wrote:
> Hi,
> On Fri, 2019-05-17 at 16:05 +0300, Heikki Krogerus wrote:
> > Hi,
> > 
> > On Fri, May 17, 2019 at 01:37:36PM +0300, Heikki Krogerus wrote:
> > > On Tue, May 14, 2019 at 04:47:21PM +0800, Chunfeng Yun wrote:
> > > > Add fwnode_usb_role_switch_get() to make easier to get
> > > > usb_role_switch by fwnode which register it.
> > > > It's useful when there is not device_connection registered
> > > > between two drivers and only knows the fwnode which register
> > > > usb_role_switch.
> > > > 
> > > > Signed-off-by: Chunfeng Yun <chunfeng.yun@xxxxxxxxxxxx>
> > > > Tested-by: Biju Das <biju.das@xxxxxxxxxxxxxx>
> > > 
> > > Acked-by: Heikki Krogerus <heikki.krogerus@xxxxxxxxxxxxxxx>
> > 
> > Hold on. I just noticed Rob's comment on patch 2/6, where he points out
> > that you don't need to use device graph since the controller is the
> > parent of the connector. Doesn't that mean you don't really need this
> > API?
> No, I still need it. 
> The change is about the way how to get fwnode;
> when use device graph, get fwnode by of_graph_get_remote_node();
> but now will get fwnode by of_get_parent();

OK, I get that, but I'm still not convinced about if something like
this function is needed at all. I also have concerns regarding how you
are using the function. I'll explain in comment to the patch 5/6 in
this series...

> > > > ---
> > > > v5 changes:
> > > >  1. remove linux/of.h suggested by Biju
> > > >  2. add tested by Biju
> > > > 
> > > > Note: still depends on [1]
> > > >  [1]: [v6,08/13] usb: roles: Introduce stubs for the exiting functions in role.h
> > > >       https://patchwork.kernel.org/patch/10909971/
> > > > 
> > > > v4 changes:
> > > >   1. use switch_fwnode_match() to find fwnode suggested by Heikki
> > > >   2. this patch now depends on [1]
> > > > 
> > > >  [1] [v6,08/13] usb: roles: Introduce stubs for the exiting functions in role.h
> > > >     https://patchwork.kernel.org/patch/10909971/
> > > > 
> > > > v3 changes:
> > > >   1. use fwnodes instead of node suggested by Andy
> > > >   2. rebuild the API suggested by Heikki
> > > > 
> > > > v2 no changes
> > > > ---
> > > >  drivers/usb/roles/class.c | 24 ++++++++++++++++++++++++
> > > >  include/linux/usb/role.h  |  8 ++++++++
> > > >  2 files changed, 32 insertions(+)
> > > > 
> > > > diff --git a/drivers/usb/roles/class.c b/drivers/usb/roles/class.c
> > > > index f45d8df5cfb8..4a1f09a41ec0 100644
> > > > --- a/drivers/usb/roles/class.c
> > > > +++ b/drivers/usb/roles/class.c
> > > > @@ -135,6 +135,30 @@ struct usb_role_switch *usb_role_switch_get(struct device *dev)
> > > >  }
> > > >  EXPORT_SYMBOL_GPL(usb_role_switch_get);
> > > >  
> > > > +/**
> > > > + * fwnode_usb_role_switch_get - Find USB role switch by it's parent fwnode
> > > > + * @fwnode: The fwnode that register USB role switch
> > > > + *
> > > > + * Finds and returns role switch registered by @fwnode. The reference count
> > > > + * for the found switch is incremented.
> > > > + */
> > > > +struct usb_role_switch *
> > > > +fwnode_usb_role_switch_get(struct fwnode_handle *fwnode)
> > > > +{
> > > > +	struct usb_role_switch *sw;
> > > > +	struct device *dev;
> > > > +
> > > > +	dev = class_find_device(role_class, NULL, fwnode, switch_fwnode_match);
> > > > +	if (!dev)
> > > > +		return ERR_PTR(-EPROBE_DEFER);
> > > > +
> > > > +	sw = to_role_switch(dev);
> > > > +	WARN_ON(!try_module_get(sw->dev.parent->driver->owner));
> > > > +
> > > > +	return sw;
> > > > +}
> > > > +EXPORT_SYMBOL_GPL(fwnode_usb_role_switch_get);
> > 
> > This function only basically converts the fwnode to usb_role_switch,
> > but I would actually prefer that we walked through the device graph
> > here instead of expecting the caller to do that.
> > 
> > So this function should probable be called fwnode_to_usb_role_switch()
> > and not fwnode_usb_role_switch_get(), but I guess you don't need it
> > at all, right?
> > 
> > 
> > thanks,
> > 
> 

-- 
heikki