Re: [PATCH v2 05/12] net: stmmac: dwmac-starfive: Add support for JH7100 SoC

On 10/31/23 16:33, Emil Renner Berthing wrote:

Cristian Ciocaltea wrote:

quoted

Add a missing quirk to enable support for the StarFive JH7100 SoC.

Additionally, for greater flexibility in operation, allow using the
rgmii-rxid and rgmii-txid phy modes.

Co-developed-by: Emil Renner Berthing <kernel@esmil.dk>
Signed-off-by: Emil Renner Berthing <kernel@esmil.dk>
Signed-off-by: Cristian Ciocaltea <redacted>

Hi Cristian,

Thanks for working on this! This driver has code to update the phy clock for
different line speeds. I don't think that will work without the
CLK_SET_RATE_PARENT flag added to the clock in [1] which in turn depends on
[2].

[1]: https://github.com/esmil/linux/commit/b200c3054b58a49ba25af67aff82d9045e3c3666
[2]: https://github.com/esmil/linux/commit/dce189542c16bf0eb8533d96c0305cb59d149dae

Two more comments below..

Hi Emil,

Thanks for the review!

I've been only testing this at 1000 Mbps and so far it seems to work
fine. I will try with different line speeds and report back.

quoted

---
 drivers/net/ethernet/stmicro/stmmac/Kconfig   |  6 ++--
 .../ethernet/stmicro/stmmac/dwmac-starfive.c  | 32 ++++++++++++++++---
 2 files changed, 31 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/stmicro/stmmac/Kconfig b/drivers/net/ethernet/stmicro/stmmac/Kconfig
index a2b9e289aa36..c3c2c8360047 100644
--- a/drivers/net/ethernet/stmicro/stmmac/Kconfig
+++ b/drivers/net/ethernet/stmicro/stmmac/Kconfig

@@ -165,9 +165,9 @@ config DWMAC_STARFIVE
 	help
 	  Support for ethernet controllers on StarFive RISC-V SoCs

-	  This selects the StarFive platform specific glue layer support for
-	  the stmmac device driver. This driver is used for StarFive JH7110
-	  ethernet controller.
+	  This selects the StarFive platform specific glue layer support
+	  for the stmmac device driver. This driver is used for the
+	  StarFive JH7100 and JH7110 ethernet controllers.

 config DWMAC_STI
 	tristate "STi GMAC support"

diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-starfive.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-starfive.c
index 5d630affb4d1..88c431edcea0 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac-starfive.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-starfive.c

@@ -15,13 +15,20 @@

 #include "stmmac_platform.h"

-#define STARFIVE_DWMAC_PHY_INFT_RGMII	0x1
-#define STARFIVE_DWMAC_PHY_INFT_RMII	0x4
-#define STARFIVE_DWMAC_PHY_INFT_FIELD	0x7U
+#define STARFIVE_DWMAC_PHY_INFT_RGMII		0x1
+#define STARFIVE_DWMAC_PHY_INFT_RMII		0x4
+#define STARFIVE_DWMAC_PHY_INFT_FIELD		0x7U
+
+#define JH7100_SYSMAIN_REGISTER49_DLYCHAIN	0xc8
+
+struct starfive_dwmac_data {
+	unsigned int gtxclk_dlychain;
+};

 struct starfive_dwmac {
 	struct device *dev;
 	struct clk *clk_tx;
+	const struct starfive_dwmac_data *data;
 };

 static void starfive_dwmac_fix_mac_speed(void *priv, unsigned int speed, unsigned int mode)

@@ -67,6 +74,8 @@ static int starfive_dwmac_set_mode(struct plat_stmmacenet_data *plat_dat)

 	case PHY_INTERFACE_MODE_RGMII:
 	case PHY_INTERFACE_MODE_RGMII_ID:
+	case PHY_INTERFACE_MODE_RGMII_RXID:
+	case PHY_INTERFACE_MODE_RGMII_TXID:
 		mode = STARFIVE_DWMAC_PHY_INFT_RGMII;
 		break;

@@ -89,6 +98,14 @@ static int starfive_dwmac_set_mode(struct plat_stmmacenet_data *plat_dat)
 	if (err)
 		return dev_err_probe(dwmac->dev, err, "error setting phy mode\n");

+	if (dwmac->data) {

I think you want something like this so future quirks don't need to touch this
code:

	if (dwmac->data && dwmac->data->gtxclk_dlychain)

Yes, but that would prevent having a quirk that explicitly wants to write 0.

I was initially thinking to something more generic, like providing a
list of register-value pairs, but I'm not sure this is going to be ever
needed.

I'm still open to apply the suggested change, though.

quoted

+		err = regmap_write(regmap, JH7100_SYSMAIN_REGISTER49_DLYCHAIN,
+				   dwmac->data->gtxclk_dlychain);
+		if (err)
+			return dev_err_probe(dwmac->dev, err,
+					     "error selecting gtxclk delay chain\n");
+	}
+
 	return 0;
 }

@@ -114,6 +131,8 @@ static int starfive_dwmac_probe(struct platform_device *pdev)
 	if (!dwmac)
 		return -ENOMEM;

+	dwmac->data = device_get_match_data(&pdev->dev);
+
 	dwmac->clk_tx = devm_clk_get_enabled(&pdev->dev, "tx");
 	if (IS_ERR(dwmac->clk_tx))
 		return dev_err_probe(&pdev->dev, PTR_ERR(dwmac->clk_tx),

@@ -144,8 +163,13 @@ static int starfive_dwmac_probe(struct platform_device *pdev)
 	return stmmac_dvr_probe(&pdev->dev, plat_dat, &stmmac_res);
 }

+static const struct starfive_dwmac_data jh7100_data = {
+	.gtxclk_dlychain = 4

Please add a , at the end of this line. I know it's unlikely that we need to
add more properties, but it's still good practice to do. This way such patches
won't need to touch this line.

Sure, will do.

quoted

+};
+
 static const struct of_device_id starfive_dwmac_match[] = {
-	{ .compatible = "starfive,jh7110-dwmac"	},
+	{ .compatible = "starfive,jh7100-dwmac", .data = &jh7100_data },
+	{ .compatible = "starfive,jh7110-dwmac" },
 	{ /* sentinel */ }
 };
 MODULE_DEVICE_TABLE(of, starfive_dwmac_match);
--
2.42.0