Re: [PATCH RFC net-next 12/13] net: phylink: add struct phylink_pcs

[Vladimir Oltean <olteanv@gmail.com>]

On Wed, 1 Jul 2020 at 14:16, Russell King - ARM Linux admin
<linux@armlinux.org.uk> wrote:
>
> On Wed, Jul 01, 2020 at 01:47:27PM +0300, Vladimir Oltean wrote:
> > Hi Russell,
> >
> > On Tue, 30 Jun 2020 at 17:29, Russell King <rmk+kernel@armlinux.org.uk> wrote:
> > >
> > > Add a way for MAC PCS to have private data while keeping independence
> > > from struct phylink_config, which is used for the MAC itself. We need
> > > this independence as we will have stand-alone code for PCS that is
> > > independent of the MAC.  Introduce struct phylink_pcs, which is
> > > designed to be embedded in a driver private data structure.
> > >
> > > This structure does not include a mdio_device as there are PCS
> > > implementations such as the Marvell DSA and network drivers where this
> > > is not necessary.
> > >
> > > Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
> > > ---
> > >  drivers/net/phy/phylink.c | 25 ++++++++++++++++------
> > >  include/linux/phylink.h   | 45 ++++++++++++++++++++++++++-------------
> > >  2 files changed, 48 insertions(+), 22 deletions(-)
> > >
> > > diff --git a/drivers/net/phy/phylink.c b/drivers/net/phy/phylink.c
> > > index a31a00fb4974..fbc8591b474b 100644
> > > --- a/drivers/net/phy/phylink.c
> > > +++ b/drivers/net/phy/phylink.c
> > > @@ -43,6 +43,7 @@ struct phylink {
> > >         const struct phylink_mac_ops *mac_ops;
> > >         const struct phylink_pcs_ops *pcs_ops;
> > >         struct phylink_config *config;
> > > +       struct phylink_pcs *pcs;
> > >         struct device *dev;
> > >         unsigned int old_link_state:1;
> > >
> > > @@ -427,7 +428,7 @@ static void phylink_mac_pcs_an_restart(struct phylink *pl)
> > >             phy_interface_mode_is_8023z(pl->link_config.interface) &&
> > >             phylink_autoneg_inband(pl->cur_link_an_mode)) {
> > >                 if (pl->pcs_ops)
> > > -                       pl->pcs_ops->pcs_an_restart(pl->config);
> > > +                       pl->pcs_ops->pcs_an_restart(pl->pcs);
> > >                 else
> > >                         pl->mac_ops->mac_an_restart(pl->config);
> > >         }
> > > @@ -453,7 +454,7 @@ static void phylink_change_interface(struct phylink *pl, bool restart,
> > >         phylink_mac_config(pl, state);
> > >
> > >         if (pl->pcs_ops) {
> > > -               err = pl->pcs_ops->pcs_config(pl->config, pl->cur_link_an_mode,
> > > +               err = pl->pcs_ops->pcs_config(pl->pcs, pl->cur_link_an_mode,
> > >                                               state->interface,
> > >                                               state->advertising,
> > >                                               !!(pl->link_config.pause &
> > > @@ -533,7 +534,7 @@ static void phylink_mac_pcs_get_state(struct phylink *pl,
> > >         state->link = 1;
> > >
> > >         if (pl->pcs_ops)
> > > -               pl->pcs_ops->pcs_get_state(pl->config, state);
> > > +               pl->pcs_ops->pcs_get_state(pl->pcs, state);
> > >         else
> > >                 pl->mac_ops->mac_pcs_get_state(pl->config, state);
> > >  }
> > > @@ -604,7 +605,7 @@ static void phylink_link_up(struct phylink *pl,
> > >         pl->cur_interface = link_state.interface;
> > >
> > >         if (pl->pcs_ops && pl->pcs_ops->pcs_link_up)
> > > -               pl->pcs_ops->pcs_link_up(pl->config, pl->cur_link_an_mode,
> > > +               pl->pcs_ops->pcs_link_up(pl->pcs, pl->cur_link_an_mode,
> > >                                          pl->cur_interface,
> > >                                          link_state.speed, link_state.duplex);
> > >
> > > @@ -863,11 +864,19 @@ struct phylink *phylink_create(struct phylink_config *config,
> > >  }
> > >  EXPORT_SYMBOL_GPL(phylink_create);
> > >
> > > -void phylink_add_pcs(struct phylink *pl, const struct phylink_pcs_ops *ops)
> > > +/**
> > > + * phylink_set_pcs() - set the current PCS for phylink to use
> > > + * @pl: a pointer to a &struct phylink returned from phylink_create()
> > > + * @pcs: a pointer to the &struct phylink_pcs
> > > + *
> > > + * Bind the MAC PCS to phylink.
> > > + */
> > > +void phylink_set_pcs(struct phylink *pl, struct phylink_pcs *pcs)
> > >  {
> > > -       pl->pcs_ops = ops;
> > > +       pl->pcs = pcs;
> > > +       pl->pcs_ops = pcs->ops;
> > >  }
> > > -EXPORT_SYMBOL_GPL(phylink_add_pcs);
> > > +EXPORT_SYMBOL_GPL(phylink_set_pcs);
> > >
> > >  /**
> > >   * phylink_destroy() - cleanup and destroy the phylink instance
> > > @@ -1212,6 +1221,8 @@ void phylink_start(struct phylink *pl)
> > >                 break;
> > >         case MLO_AN_INBAND:
> > >                 poll |= pl->config->pcs_poll;
> > > +               if (pl->pcs)
> > > +                       poll |= pl->pcs->poll;
> >
> > Do we see a need for yet another way to request phylink to poll the
> > PCS for link status?
>
> Please consider what the model looks like if we have the PCS almost
> self contained except for this property, which is in the MAC side.
> What if some PCS need to be polled but others do not.  Why should the
> MAC need to have that knowledge - is it not a property of the PCS
> itself?
>
> The reason we stuffed it into phylink_config is that at the time, that
> was all that existed.  That doesn't mean that when we change the model
> that we should be tied by that decision.
>
> So, for example, does the Lynx PCS IP support any kind of notification
> of link changes to its integrated system?  If it does not, then having
> the Lynx PCS mark _itself_ as needing polling is entirely sane, rather
> than burying that information in the MAC driver.
>
> --
> RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
> FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!

Well, there is a MAC register called IEVENT[PCS]. For some Lynx
integrations, such as Felix, I know it doesn't fire an IRQ. Now, if it
doesn't fire on _all_ SoCs which integrate Lynx, that I don't know.
But the interrupt is going to be highly system-specific either way (on
some MACs it's a regular IRQ line, on others it's an MSI). So, in the
general case, I think this is system-specific and not a property of
the PCS itself.