RE: [PATCH 2/2] net: dsa: b53: mdio: add support for BCM53134

[Paul Geurts]

> -----Original Message-----
> From: Álvaro Fernández Rojas <noltari@xxxxxxxxx>
> Sent: donderdag 23 maart 2023 23:13
> To: Florian Fainelli <f.fainelli@xxxxxxxxx>
> Cc: Paul Geurts <paul.geurts@xxxxxxxxxxxxxxxxxxxxxxxxx>;
> jonas.gorski@xxxxxxxxx; andrew@xxxxxxx; olteanv@xxxxxxxxx;
> davem@xxxxxxxxxxxxx; edumazet@xxxxxxxxxx; kuba@xxxxxxxxxx;
> pabeni@xxxxxxxxxx; robh+dt@xxxxxxxxxx;
> krzysztof.kozlowski+dt@xxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 2/2] net: dsa: b53: mdio: add support for BCM53134
> 
> El jue, 23 mar 2023 a las 22:02, Florian Fainelli
> (<f.fainelli@xxxxxxxxx>) escribió:
> >
> > On 3/23/23 13:10, Paul Geurts wrote:
> > >> -----Original Message-----
> > >> From: Florian Fainelli <f.fainelli@xxxxxxxxx>
> > >> Sent: donderdag 23 maart 2023 17:43
> > >> To: Álvaro Fernández Rojas <noltari@xxxxxxxxx>; Paul Geurts
> > >> <paul.geurts@xxxxxxxxxxxxxxxxxxxxxxxxx>; jonas.gorski@xxxxxxxxx;
> > >> andrew@xxxxxxx; olteanv@xxxxxxxxx; davem@xxxxxxxxxxxxx;
> > >> edumazet@xxxxxxxxxx; kuba@xxxxxxxxxx; pabeni@xxxxxxxxxx;
> > >> robh+dt@xxxxxxxxxx; krzysztof.kozlowski+dt@xxxxxxxxxx;
> > >> netdev@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-
> > >> kernel@xxxxxxxxxxxxxxx
> > >> Subject: Re: [PATCH 2/2] net: dsa: b53: mdio: add support for
> > >> BCM53134
> > >>
> > >> On 3/23/23 05:18, Álvaro Fernández Rojas wrote:
> > >>> From: Paul Geurts <paul.geurts@xxxxxxxxxxxxxxxxxxxxxxxxx>
> > >>>
> > >>> Add support for the BCM53134 Ethernet switch in the existing b53
> > >>> dsa
> > >> driver.
> > >>> BCM53134 is very similar to the BCM58XX series.
> > >>>
> > >>> Signed-off-by: Paul Geurts <paul.geurts@xxxxxxxxxxxxxxxxxxxxxxxxx>
> > >>> Signed-off-by: Álvaro Fernández Rojas <noltari@xxxxxxxxx>
> > >>> ---
> > >>>    drivers/net/dsa/b53/b53_common.c | 53
> > >> +++++++++++++++++++++++++++++++-
> > >>>    drivers/net/dsa/b53/b53_mdio.c   |  5 ++-
> > >>>    drivers/net/dsa/b53/b53_priv.h   |  9 +++++-
> > >>>    3 files changed, 64 insertions(+), 3 deletions(-)
> > >>>
> > >>> diff --git a/drivers/net/dsa/b53/b53_common.c
> > >>> b/drivers/net/dsa/b53/b53_common.c
> > >>> index 1f9b251a5452..aaa0813e6f59 100644
> > >>> --- a/drivers/net/dsa/b53/b53_common.c
> > >>> +++ b/drivers/net/dsa/b53/b53_common.c
> > >>> @@ -1282,6 +1282,42 @@ static void b53_adjust_link(struct
> > >>> dsa_switch
> > >> *ds, int port,
> > >>>     if (is63xx(dev) && port >= B53_63XX_RGMII0)
> > >>>             b53_adjust_63xx_rgmii(ds, port, phydev->interface);
> > >>>
> > >>> +   if (is53134(dev) && phy_interface_is_rgmii(phydev)) {
> > >>
> > >> Why is not this in the same code block as the one for the is531x5()
> > >> device like
> > >> this:
> > >>
> > >> diff --git a/drivers/net/dsa/b53/b53_common.c
> > >> b/drivers/net/dsa/b53/b53_common.c
> > >> index 59cdfc51ce06..1c64b6ce7e78 100644
> > >> --- a/drivers/net/dsa/b53/b53_common.c
> > >> +++ b/drivers/net/dsa/b53/b53_common.c
> > >> @@ -1235,7 +1235,7 @@ static void b53_adjust_link(struct dsa_switch
> > >> *ds, int port,
> > >>                                 tx_pause, rx_pause);
> > >>           b53_force_link(dev, port, phydev->link);
> > >>
> > >> -       if (is531x5(dev) && phy_interface_is_rgmii(phydev)) {
> > >> +       if ((is531x5(dev) || is53134(dev)) &&
> > >> phy_interface_is_rgmii(phydev)) {
> > >>                   if (port == dev->imp_port)
> > >>                           off = B53_RGMII_CTRL_IMP;
> > >>                   else
> > >>
> > >> Other than that, LGTM!
> > >> --
> > >> Florian
> > >
> > > I think the only reason is that the BCM53134 does not support the
> > > RGMII_CTRL_TIMING_SEL bit, which is set in the original block. I
> > > agree Putting a if statement around rgmii_ctrl |=
> > > RGMII_CTRL_TIMING_SEL; would prevent a lot of code duplication.
> > > _however_, after looking at it again, I don’t think the device does
> > > not support the bit. When looking at the datasheet, The same bit in
> > > the this register is called BYPASS_2NS_DEL. It's very uncommon For
> > > Broadcom to make such a change in the register interface, so maybe
> > > they Just renamed it. Do you think this could be the same bit?
> >
> > Yes, I think this is exactly the same bit, just named differently.
> > What strikes me as odd is that neither of the 53115, 53125 or 53128
> > which are guarded by the is531x5() conditional have it defined.
> 
> If this is true then we can safely add 53134 to is531x5() conditional and reuse
> the existing code.
> 
> > --
> > Florian
> >
> 
> --
> Álvaro

With the bit set on my device, everything keeps working just fine. I indeed think
This is just the same bit. I have requested some clarity from our FAE at
Broadcom. We could wait for their answer (which could take a while), or just
assume setting this bit is fine an push the patches. I'll leave that up to you.

---
Paul