[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3304458.aeNJFYEL58@trenzalore>
Date: Fri, 09 Aug 2024 10:38:23 -0400
From: Detlev Casanova <detlev.casanova@...labora.com>
To: linux-kernel@...r.kernel.org,
Heiko Stübner <heiko@...ech.de>
Cc: "David S . Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>, Rob Herring <robh@...nel.org>,
Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley <conor+dt@...nel.org>,
Alexandre Torgue <alexandre.torgue@...s.st.com>,
Jose Abreu <joabreu@...opsys.com>,
Maxime Coquelin <mcoquelin.stm32@...il.com>,
David Wu <david.wu@...k-chips.com>,
Giuseppe Cavallaro <peppe.cavallaro@...com>, netdev@...r.kernel.org,
devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-rockchip@...ts.infradead.org, linux-stm32@...md-mailman.stormreply.com,
kernel@...labora.com
Subject:
Re: [PATCH v2 2/2] ethernet: stmmac: dwmac-rk: Add GMAC support for RK3576
On Friday, 9 August 2024 09:16:44 EDT Heiko Stübner wrote:
> Hi Detlev,
>
> Am Donnerstag, 8. August 2024, 19:00:18 CEST schrieb Detlev Casanova:
> > From: David Wu <david.wu@...k-chips.com>
> >
> > Add constants and callback functions for the dwmac on RK3576 soc.
> >
> > Signed-off-by: David Wu <david.wu@...k-chips.com>
> > [rebase, extracted bindings]
> > Signed-off-by: Detlev Casanova <detlev.casanova@...labora.com>
> > ---
> >
> > .../net/ethernet/stmicro/stmmac/dwmac-rk.c | 156 ++++++++++++++++++
> > 1 file changed, 156 insertions(+)
> >
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
> > b/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c index
> > 7ae04d8d291c8..e1fa8fc9f4012 100644
> > --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
> > +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
> > @@ -1116,6 +1116,161 @@ static const struct rk_gmac_ops rk3568_ops = {
> >
> > },
> >
> > };
> >
[...]
> > +/* SDGMAC_GRF */
> > +#define RK3576_GRF_GMAC_CON0 0X0020
> > +#define RK3576_GRF_GMAC_CON1 0X0024
> > +
> > +#define RK3576_GMAC_RMII_MODE GRF_BIT(3)
> > +#define RK3576_GMAC_RGMII_MODE GRF_CLR_BIT(3)
> > +
> > +#define RK3576_GMAC_CLK_SELET_IO GRF_BIT(7)
> > +#define RK3576_GMAC_CLK_SELET_CRU GRF_CLR_BIT(7)
>
> nit: typos _CLK_SELECT_ ... missing the C in select
Ack
> > +
> > +#define RK3576_GMAC_CLK_RMII_DIV2 GRF_BIT(5)
> > +#define RK3576_GMAC_CLK_RMII_DIV20 GRF_CLR_BIT(5)
>
> I think those are backwards
> The TRM says bit[5]=0: 25MHz (DIV2) and bit[5]=1: 2.5MHz (DIV20)
>
> I guess nobody also on Rockchip's side tested a RMII phy on those controllrs
Can't be sure about that. An error in the TRM is not impossible either, as for
rk3588, it is also bit[5]=0: DIV20 and bit[5]=1: DIV2. I can switch them to
match the TRM though, we may never now.
> > +
> > +#define RK3576_GMAC_CLK_RGMII_DIV1 \
> > + (GRF_CLR_BIT(6) | GRF_CLR_BIT(5))
> > +#define RK3576_GMAC_CLK_RGMII_DIV5 \
> > + (GRF_BIT(6) | GRF_BIT(5))
> > +#define RK3576_GMAC_CLK_RGMII_DIV50 \
> > + (GRF_BIT(6) | GRF_CLR_BIT(5))
> > +
>
> in contrast, these are correct and match the TRM
>
> > +#define RK3576_GMAC_CLK_RMII_GATE GRF_BIT(4)
> > +#define RK3576_GMAC_CLK_RMII_NOGATE GRF_CLR_BIT(4)
> > +
> > +static void rk3576_set_to_rgmii(struct rk_priv_data *bsp_priv,
> > + int tx_delay, int rx_delay)
> > +{
> > + struct device *dev = &bsp_priv->pdev->dev;
> > + unsigned int offset_con;
> > +
> > + if (IS_ERR(bsp_priv->grf) || IS_ERR(bsp_priv->php_grf)) {
> > + dev_err(dev, "Missing rockchip,grf or rockchip,php_grf
property\n");
> > + return;
> > + }
> > +
> > + offset_con = bsp_priv->id == 1 ? RK3576_GRF_GMAC_CON1 :
> > + RK3576_GRF_GMAC_CON0;
> > +
> > + regmap_write(bsp_priv->grf, offset_con, RK3576_GMAC_RGMII_MODE);
> > +
> > + offset_con = bsp_priv->id == 1 ? RK3576_VCCIO0_1_3_IOC_CON4 :
> > +
RK3576_VCCIO0_1_3_IOC_CON2;
> > +
> > + /* m0 && m1 delay enabled */
> > + regmap_write(bsp_priv->php_grf, offset_con,
> > + DELAY_ENABLE(RK3576, tx_delay, rx_delay));
> > + regmap_write(bsp_priv->php_grf, offset_con + 0x4,
> > + DELAY_ENABLE(RK3576, tx_delay, rx_delay));
> > +
> > + /* m0 && m1 delay value */
> > + regmap_write(bsp_priv->php_grf, offset_con,
> > + RK3576_GMAC_CLK_TX_DL_CFG(tx_delay) |
> > + RK3576_GMAC_CLK_RX_DL_CFG(rx_delay));
> > + regmap_write(bsp_priv->php_grf, offset_con + 0x4,
> > + RK3576_GMAC_CLK_TX_DL_CFG(tx_delay) |
> > + RK3576_GMAC_CLK_RX_DL_CFG(rx_delay));
> > +}
> > +
> > +static void rk3576_set_to_rmii(struct rk_priv_data *bsp_priv)
> > +{
> > + struct device *dev = &bsp_priv->pdev->dev;
> > + unsigned int offset_con;
> > +
> > + if (IS_ERR(bsp_priv->php_grf)) {
> > + dev_err(dev, "%s: Missing rockchip,php_grf property\n",
__func__);
> > + return;
> > + }
> > +
> > + offset_con = bsp_priv->id == 1 ? RK3576_GRF_GMAC_CON1 :
> > + RK3576_GRF_GMAC_CON0;
> > +
> > + regmap_write(bsp_priv->grf, offset_con, RK3576_GMAC_RMII_MODE);
> > +}
> > +
> > +static void rk3576_set_gmac_speed(struct rk_priv_data *bsp_priv, int
> > speed) +{
> > + struct device *dev = &bsp_priv->pdev->dev;
> > + unsigned int val = 0, offset_con;
> > +
> > + switch (speed) {
> > + case 10:
> > + if (bsp_priv->phy_iface == PHY_INTERFACE_MODE_RMII)
> > + val = RK3576_GMAC_CLK_RMII_DIV20;
> > + else
> > + val = RK3576_GMAC_CLK_RGMII_DIV50;
>
> val = bsp_priv->phy_iface == PHY_INTERFACE_MODE_RMII ?
> RK3576_GMAC_CLK_RMII_DIV20 :
> RK3576_GMAC_CLK_RGMII_DIV50;
> perhaps?
This way matches how it is written in rk3588_set_gmac_speed(). I find that
having similar code for similar functions helps reading and understanding it
better (although I agree that your suggestion looks better).
I'd rather keep it like it is for now if that's ok.
> > + break;
> > + case 100:
> > + if (bsp_priv->phy_iface == PHY_INTERFACE_MODE_RMII)
> > + val = RK3576_GMAC_CLK_RMII_DIV2;
> > + else
> > + val = RK3576_GMAC_CLK_RGMII_DIV5;
>
> same as above?
>
> > + break;
> > + case 1000:
> > + if (bsp_priv->phy_iface != PHY_INTERFACE_MODE_RMII)
> > + val = RK3576_GMAC_CLK_RGMII_DIV1;
> > + else
> > + goto err;
>
> if (bsp_priv->phy_iface == PHY_INTERFACE_MODE_RMII)
> goto err;
>
> val = RK3576_GMAC_CLK_RGMII_DIV1;
>
> > + break;
> > + default:
> > + goto err;
> > + }
> > +
> > + offset_con = bsp_priv->id == 1 ? RK3576_GRF_GMAC_CON1 :
> > + RK3576_GRF_GMAC_CON0;
> > +
> > + regmap_write(bsp_priv->grf, offset_con, val);
> > +
> > + return;
> > +err:
> > + dev_err(dev, "unknown speed value for GMAC speed=%d", speed);
> > +}
> > +
> > +static void rk3576_set_clock_selection(struct rk_priv_data *bsp_priv,
> > bool input, + bool enable)
> > +{
> > + unsigned int val = input ? RK3576_GMAC_CLK_SELET_IO :
> > + RK3576_GMAC_CLK_SELET_CRU;
> > + unsigned int offset_con;
> > +
> > + val |= enable ? RK3576_GMAC_CLK_RMII_NOGATE :
> > + RK3576_GMAC_CLK_RMII_GATE;
> > +
> > + offset_con = bsp_priv->id == 1 ? RK3576_GRF_GMAC_CON1 :
> > + RK3576_GRF_GMAC_CON0;
>
> nit: alignment of both looks like it could be nicer
That's strange, the alignments looks good in vim and git diff. It also looks
nice on the archive: https://lore.kernel.org/linux-rockchip/
20240808170113.82775-3-detlev.casanova@...labora.com/
Regards,
Detlev
Powered by blists - more mailing lists