Re: [PATCH v2 1/2] phy: add new phy_notify_pmstate() api

[neil . armstrong]

On 23/07/2025 09:07, Vinod Koul wrote:
> On 22-07-25, 22:04, Manivannan Sadhasivam wrote:
> > On Thu, Jul 03, 2025 at 02:03:22PM GMT, Peter Griffin wrote:
> > > Add a new phy_notify_pmstate() api that notifies and configures a phy for a
> > > given PM link state transition.
> > >
> > > This is intended to be by phy drivers which need to do some runtime
> > > configuration of parameters during the transition that can't be handled by
> > > phy_calibrate() or phy_power_{on|off}().
> > >
> > > The first usage of this API is in the Samsung UFS phy that needs to issue
> > > some register writes when entering and exiting the hibernate link state.
> > >
> > > Signed-off-by: Peter Griffin <peter.griffin@xxxxxxxxxx>
> > > ---
> > >   drivers/phy/phy-core.c  | 25 +++++++++++++++++++++++++
> > >   include/linux/phy/phy.h | 25 +++++++++++++++++++++++++
> > >   2 files changed, 50 insertions(+)
> > >
> > > diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c
> > > index 04a5a34e7a950ae94fae915673c25d476fc071c1..0b29bc2c709890d7fc27d1480a35cda6a826fd30 100644
> > > --- a/drivers/phy/phy-core.c
> > > +++ b/drivers/phy/phy-core.c
> > > @@ -520,6 +520,31 @@ int phy_notify_disconnect(struct phy *phy, int port)
> > >   }
> > >   EXPORT_SYMBOL_GPL(phy_notify_disconnect);
> > >   
> > > +/**
> > > + * phy_notify_pmstate() - phy link state notification
> >
> > 'pmstate' doesn't correspond to 'link state'. So how about,
> > phy_notify_link_state()?
> >
> > s/phy/PHY (here and below)
> >
> > > + * @phy: the phy returned by phy_get()
> > > + * @state: the link state
> > > + *
> > > + * Notify the phy of some PM link state transition. Used to notify and
> >
> > Link state change is common for the PHY. So remove 'PM'.
>
> Is it really link or phy state?

It seems to be a link state, and I think adding a way to force
a link state to a phy, which is basically the role of a phy,
more coherent.

Neil

> > > + * configure the phy accordingly.
> > > + *
> > > + * Returns: %0 if successful, a negative error code otherwise
> > > + */
> > > +int phy_notify_pmstate(struct phy *phy, enum phy_linkstate state)
> >
> > I think you need to use 'int state' and let drivers pass their own link state
> > values. You cannot have generic link states across all peripherals.
>
> I would avoid that, people start overloading this if we let it keep
> open! I would like to avoid the api -(ab)use