Date:   Thu, 7 Dec 2023 10:28:42 +0000
From:   Stanley Chang[昌育德] 
        <stanley_chang@...ltek.com>
To:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>
CC:     Vinod Koul <vkoul@...nel.org>, Johan Hovold <johan@...nel.org>,
        "Kishon Vijay Abraham I" <kishon@...nel.org>,
        Geert Uytterhoeven <geert+renesas@...der.be>,
        Jinjie Ruan <ruanjinjie@...wei.com>,
        Rob Herring <robh@...nel.org>,
        Alan Stern <stern@...land.harvard.edu>,
        Heikki Krogerus <heikki.krogerus@...ux.intel.com>,
        Flavio Suligoi <f.suligoi@...m.it>,
        Ricardo Cañuelo 
        <ricardo.canuelo@...labora.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-phy@...ts.infradead.org" <linux-phy@...ts.infradead.org>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>
Subject: RE: [PATCH v3 RESEND 1/4] phy: core: add notify_connect and notify_disconnect callback

Hi Greg,

> 
> On Thu, Dec 07, 2023 at 03:38:04PM +0800, Stanley Chang wrote:
> > In Realtek SoC, the parameter of usb phy is designed to can dynamic
> > tuning base on port status. Therefore, add a notify callback of phy
> > driver when usb connection/disconnection change.
> >
> > Signed-off-by: Stanley Chang <stanley_chang@...ltek.com>
> > ---
> > RESEND:
> >     Because there is no extcon device provided in the USB framework to
> >     notify connect and disconnect.
> >     Therefore, I added the notification connection/disconnection based
> >     on the generic phy. So I no use the EXTCON framework for notifying
> >     connect/disconnect.
> > v2 to v3:
> >     No change
> > v1 to v2:
> >     No change
> > ---
> >  drivers/phy/phy-core.c  | 47
> > +++++++++++++++++++++++++++++++++++++++++
> >  include/linux/phy/phy.h | 18 ++++++++++++++++
> >  2 files changed, 65 insertions(+)
> >
> > diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c index
> > 96a0b1e111f3..a84ad4896b7f 100644
> > --- a/drivers/phy/phy-core.c
> > +++ b/drivers/phy/phy-core.c
> > @@ -489,6 +489,53 @@ int phy_calibrate(struct phy *phy)  }
> > EXPORT_SYMBOL_GPL(phy_calibrate);
> >
> > +/**
> > + * phy_notify_connect() - phy connect notify
> > + * @phy: the phy returned by phy_get()
> > + * @port: the port index for connect
> > + *
> > + * If phy need the get connection status, the callback can be used.
> > + * Returns: %0 if successful, a negative error code otherwise  */ int
> > +phy_notify_connect(struct phy *phy, int port) {
> > +     int ret;
> > +
> > +     if (!phy || !phy->ops->connect)
> > +             return 0;
> 
> How can phy be null?
> 
> And it is not successful if connect is not valid, so why not return an error there?

This is possible. If a caller no use the generic phy or no define the connect callback.
And if no define connect callback, it will do nothing. So no any error.

> > +
> > +     mutex_lock(&phy->mutex);
> > +     ret = phy->ops->connect(phy, port);
> > +     mutex_unlock(&phy->mutex);
> > +
> > +     return ret;
> > +}
> > +EXPORT_SYMBOL_GPL(phy_notify_connect);
> > +
> > +/**
> > + * phy_notify_disconnect() - phy disconnect notify
> > + * @phy: the phy returned by phy_get()
> > + * @port: the port index for disconnect
> > + *
> > + * If phy need the get disconnection status, the callback can be used.
> > + *
> > + * Returns: %0 if successful, a negative error code otherwise  */ int
> > +phy_notify_disconnect(struct phy *phy, int port) {
> > +     int ret;
> > +
> > +     if (!phy || !phy->ops->disconnect)
> > +             return 0;
> 
> Same as above.

Same as connect callback.

> > +
> > +     mutex_lock(&phy->mutex);
> > +     ret = phy->ops->disconnect(phy, port);
> > +     mutex_unlock(&phy->mutex);
> > +
> > +     return ret;
> > +}
> > +EXPORT_SYMBOL_GPL(phy_notify_disconnect);
> > +
> >  /**
> >   * phy_configure() - Changes the phy parameters
> >   * @phy: the phy returned by phy_get() diff --git
> > a/include/linux/phy/phy.h b/include/linux/phy/phy.h index
> > f6d607ef0e80..cf98cb29ddaa 100644
> > --- a/include/linux/phy/phy.h
> > +++ b/include/linux/phy/phy.h
> > @@ -122,6 +122,8 @@ struct phy_ops {
> >                           union phy_configure_opts *opts);
> >       int     (*reset)(struct phy *phy);
> >       int     (*calibrate)(struct phy *phy);
> > +     int     (*connect)(struct phy *phy, int port);
> > +     int     (*disconnect)(struct phy *phy, int port);
> 
> You forgot to document these and would have a warning from the
> documentation build if this was applied :(
> 
I will add a description.

> >       void    (*release)(struct phy *phy);
> >       struct module *owner;
> >  };
> > @@ -243,6 +245,8 @@ static inline enum phy_mode phy_get_mode(struct
> > phy *phy)  }  int phy_reset(struct phy *phy);  int
> > phy_calibrate(struct phy *phy);
> > +int phy_notify_connect(struct phy *phy, int port); int
> > +phy_notify_disconnect(struct phy *phy, int port);
> >  static inline int phy_get_bus_width(struct phy *phy)  {
> >       return phy->attrs.bus_width;
> > @@ -396,6 +400,20 @@ static inline int phy_calibrate(struct phy *phy)
> >       return -ENOSYS;
> >  }
> >
> > +static inline int phy_notify_connect(struct phy *phy, int index) {
> > +     if (!phy)
> > +             return 0;
> 
> Why check this?

It is normal for phy to be null.
But if phy is not empty, -ENOSYS should be returned to indicate the error "Operation not supported".
Other callbacks also follow this rule.

> > +     return -ENOSYS;
> > +}
> > +
> > +static inline int phy_notify_disconnect(struct phy *phy, int index) {
> > +     if (!phy)
> > +             return 0;
> 
> Again, why check this?

As above.

Thanks,
Stanley

Powered by blists - more mailing lists