[PATCH v2 0/6] ARM: sunxi: Clean up sun7i-a20-gmac-clk usage

[PATCH v2 0/6] ARM: sunxi: Clean up sun7i-a20-gmac-clk usage

[Priit Laes]

This serie implements syscon-based regmap access to dwmac-sunxi driver,
allowing to deprecate the allwinner,sun7i-a20-gmac-clk clock driver.

In order to register regmap, we firstly need to change existing clock
setups (sun7i-a20 and sun6i-a31) from CLK_OF_DECLARE to
CLK_OF_DECLARE_DRIVER to allow probing the same driver again via the
platform driver framework.

This patchset touches 3 areas:
- sun7i and sun6i CCUs now set up regmap to allow dwmac-sunxi driver
to access GMAC clock register.
- dwmac-sunxi can now handle syscon-based clock register to handle
its own clock.
- sun7i and sun6i devicetrees are converted to use the new syscon-based
access.

After this patchset:
- sun7i-a20 works fully without legacy sunxi clocks (CLK_SUNXI)
- only sun9i-a80 SoC remains as a single sun7i-a20-gmac-clk user.

Please note that sun6i-a31 changes are only build-tested, as I lack
the hardware to test this.

Changes since v2:
* Fix broken sun6i-a31 CCU patch.
* Rename series to "Clean up sun7i-a20-gmac-clk usage"

Changes since v1:
* Use CLK_OF_DECLARE_DRIVER to make it possible to probe again and set up
regmap using platform device probe.
* Clarify the meaning of "legacy" in dwmac-sunxi driver.
* Make sure we don't mess with the RX/TX delay settings when updating
clock registers.
* Update devicetree bindings
* Add sun6i-A31 support. (not tested due to lack of hardware)

Priit Laes (6):
  clk: sunxi-ng: a20: Register regmap for sun7i CCU
  clk: sunxi-ng: a31: Register regmap for sun6i CCU
  net: stmmac: dwmac-sunxi: Implement syscon-based clock handling
  dt-bindings: net: sun7i-gmac: Add syscon support
  ARM: dts: sun7i: Use syscon-based implementation for gmac
  ARM: dts: sun6i: Use syscon-based implementation for gmac

 .../net/allwinner,sun7i-a20-gmac.yaml         |  15 +-
 arch/arm/boot/dts/sun6i-a31.dtsi              |  35 +----
 arch/arm/boot/dts/sun7i-a20.dtsi              |  36 +----
 drivers/clk/sunxi-ng/ccu-sun4i-a10.c          |  62 ++++++++-
 drivers/clk/sunxi-ng/ccu-sun6i-a31.c          |  62 ++++++++-
 .../net/ethernet/stmicro/stmmac/dwmac-sunxi.c | 130 ++++++++++++++++--
 6 files changed, 261 insertions(+), 79 deletions(-)

-- 
2.26.2

[PATCH v3 1/6] clk: sunxi-ng: a20: Register regmap for sun7i CCU

[Priit Laes]

On sun7i, the gmac clock is handled by the dwmac-sunxi driver, but
its configuration register is located in the CCU register range,
requiring proper regmap setup.

In order to do that, we use CLK_OF_DECLARE_DRIVER to initialize
sun7i ccu, which clears the OF_POPULATED flag, allowing the
platform device to probe the same resource with proper device node.

Signed-off-by: Priit Laes <plaes@plaes.org>
---
 drivers/clk/sunxi-ng/ccu-sun4i-a10.c | 62 +++++++++++++++++++++++++++-
 1 file changed, 60 insertions(+), 2 deletions(-)

diff --git a/drivers/clk/sunxi-ng/ccu-sun4i-a10.c b/drivers/clk/sunxi-ng/ccu-sun4i-a10.c
index f32366d9336e..fa147b8ce705 100644
--- a/drivers/clk/sunxi-ng/ccu-sun4i-a10.c
+++ b/drivers/clk/sunxi-ng/ccu-sun4i-a10.c
@@ -8,6 +8,8 @@
 #include <linux/clk-provider.h>
 #include <linux/io.h>
 #include <linux/of_address.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
 
 #include "ccu_common.h"
 #include "ccu_reset.h"
@@ -1478,5 +1480,61 @@ static void __init sun7i_a20_ccu_setup(struct device_node *node)
 {
 	sun4i_ccu_init(node, &sun7i_a20_ccu_desc);
 }
-CLK_OF_DECLARE(sun7i_a20_ccu, "allwinner,sun7i-a20-ccu",
-	       sun7i_a20_ccu_setup);
+CLK_OF_DECLARE_DRIVER(sun7i_a20_ccu, "allwinner,sun7i-a20-ccu",
+		      sun7i_a20_ccu_setup);
+
+/*
+ * Regmap for the GMAC driver (dwmac-sunxi) to allow access to
+ * GMAC configuration register.
+ */
+#define SUN7I_A20_GMAC_CFG_REG 0x164
+static bool sun7i_a20_ccu_regmap_accessible_reg(struct device *dev,
+						unsigned int reg)
+{
+	if (reg == SUN7I_A20_GMAC_CFG_REG)
+		return true;
+	return false;
+}
+
+static struct regmap_config sun7i_a20_ccu_regmap_config = {
+	.reg_bits	= 32,
+	.val_bits	= 32,
+	.reg_stride	= 4,
+	.max_register	= 0x1f4, /* clk_out_b */
+
+	.readable_reg	= sun7i_a20_ccu_regmap_accessible_reg,
+	.writeable_reg	= sun7i_a20_ccu_regmap_accessible_reg,
+};
+
+static int sun7i_a20_ccu_probe_regmap(struct platform_device *pdev)
+{
+	void __iomem *reg;
+	struct resource *res;
+	struct regmap *regmap;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	reg = devm_ioremap(&pdev->dev, res->start, resource_size(res));
+	if (IS_ERR(reg))
+		return PTR_ERR(reg);
+
+	regmap = devm_regmap_init_mmio(&pdev->dev, reg,
+				       &sun7i_a20_ccu_regmap_config);
+	if (IS_ERR(regmap))
+		return PTR_ERR(regmap);
+
+	return 0;
+}
+
+static const struct of_device_id sun7i_a20_ccu_ids[] = {
+	{ .compatible = "allwinner,sun7i-a20-ccu"},
+	{ }
+};
+
+static struct platform_driver sun7i_a20_ccu_driver = {
+	.probe = sun7i_a20_ccu_probe_regmap,
+	.driver = {
+		.name = "sun7i-a20-ccu",
+		.of_match_table = sun7i_a20_ccu_ids,
+	},
+};
+builtin_platform_driver(sun7i_a20_ccu_driver);
-- 
2.26.2

Re: [PATCH v3 1/6] clk: sunxi-ng: a20: Register regmap for sun7i CCU

[Maxime Ripard]

[-- Attachment #1: Type: text/plain, Size: 2745 bytes --]

On Thu, Apr 30, 2020 at 02:56:57PM +0300, Priit Laes wrote:
> On sun7i, the gmac clock is handled by the dwmac-sunxi driver, but
> its configuration register is located in the CCU register range,
> requiring proper regmap setup.
> 
> In order to do that, we use CLK_OF_DECLARE_DRIVER to initialize
> sun7i ccu, which clears the OF_POPULATED flag, allowing the
> platform device to probe the same resource with proper device node.
> 
> Signed-off-by: Priit Laes <plaes@plaes.org>
> ---
>  drivers/clk/sunxi-ng/ccu-sun4i-a10.c | 62 +++++++++++++++++++++++++++-
>  1 file changed, 60 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/clk/sunxi-ng/ccu-sun4i-a10.c b/drivers/clk/sunxi-ng/ccu-sun4i-a10.c
> index f32366d9336e..fa147b8ce705 100644
> --- a/drivers/clk/sunxi-ng/ccu-sun4i-a10.c
> +++ b/drivers/clk/sunxi-ng/ccu-sun4i-a10.c
> @@ -8,6 +8,8 @@
>  #include <linux/clk-provider.h>
>  #include <linux/io.h>
>  #include <linux/of_address.h>
> +#include <linux/platform_device.h>
> +#include <linux/regmap.h>
>  
>  #include "ccu_common.h"
>  #include "ccu_reset.h"
> @@ -1478,5 +1480,61 @@ static void __init sun7i_a20_ccu_setup(struct device_node *node)
>  {
>  	sun4i_ccu_init(node, &sun7i_a20_ccu_desc);
>  }
> -CLK_OF_DECLARE(sun7i_a20_ccu, "allwinner,sun7i-a20-ccu",
> -	       sun7i_a20_ccu_setup);
> +CLK_OF_DECLARE_DRIVER(sun7i_a20_ccu, "allwinner,sun7i-a20-ccu",
> +		      sun7i_a20_ccu_setup);
> +
> +/*
> + * Regmap for the GMAC driver (dwmac-sunxi) to allow access to
> + * GMAC configuration register.
> + */
> +#define SUN7I_A20_GMAC_CFG_REG 0x164
> +static bool sun7i_a20_ccu_regmap_accessible_reg(struct device *dev,
> +						unsigned int reg)
> +{
> +	if (reg == SUN7I_A20_GMAC_CFG_REG)
> +		return true;
> +	return false;
> +}
> +
> +static struct regmap_config sun7i_a20_ccu_regmap_config = {
> +	.reg_bits	= 32,
> +	.val_bits	= 32,
> +	.reg_stride	= 4,
> +	.max_register	= 0x1f4, /* clk_out_b */
> +
> +	.readable_reg	= sun7i_a20_ccu_regmap_accessible_reg,
> +	.writeable_reg	= sun7i_a20_ccu_regmap_accessible_reg,
> +};
> +
> +static int sun7i_a20_ccu_probe_regmap(struct platform_device *pdev)
> +{
> +	void __iomem *reg;
> +	struct resource *res;
> +	struct regmap *regmap;
> +
> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> +	reg = devm_ioremap(&pdev->dev, res->start, resource_size(res));
> +	if (IS_ERR(reg))
> +		return PTR_ERR(reg);

You shouldn't really create a second mapping here but reuse the one you got in
sun7i_a20_ccu_setup, since that code expect to be the sole user of it.

Storing the virtual address in a global variable should work fine since we only
ever have a single instance of the controller

Maxime

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[PATCH v3 2/6] clk: sunxi-ng: a31: Register regmap for sun6i CCU

[Priit Laes]

On sun6i, the gmac clock is handled by the dwmac-sunxi driver, but
its configuration register is located in the CCU register range,
requiring proper regmap setup.

In order to do that, we use CLK_OF_DECLARE_DRIVER to initialize
sun7i ccu, which clears the OF_POPULATED flag, allowing the
platform device to probe the same resource with device node.

Signed-off-by: Priit Laes <plaes@plaes.org>
---
 drivers/clk/sunxi-ng/ccu-sun6i-a31.c | 62 +++++++++++++++++++++++++++-
 1 file changed, 60 insertions(+), 2 deletions(-)

diff --git a/drivers/clk/sunxi-ng/ccu-sun6i-a31.c b/drivers/clk/sunxi-ng/ccu-sun6i-a31.c
index 9b40d53266a3..3f6f9824b2ca 100644
--- a/drivers/clk/sunxi-ng/ccu-sun6i-a31.c
+++ b/drivers/clk/sunxi-ng/ccu-sun6i-a31.c
@@ -10,6 +10,8 @@
 #include <linux/clk-provider.h>
 #include <linux/io.h>
 #include <linux/of_address.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
 
 #include "ccu_common.h"
 #include "ccu_reset.h"
@@ -1262,5 +1264,61 @@ static void __init sun6i_a31_ccu_setup(struct device_node *node)
 	ccu_mux_notifier_register(pll_cpu_clk.common.hw.clk,
 				  &sun6i_a31_cpu_nb);
 }
-CLK_OF_DECLARE(sun6i_a31_ccu, "allwinner,sun6i-a31-ccu",
-	       sun6i_a31_ccu_setup);
+CLK_OF_DECLARE_DRIVER(sun6i_a31_ccu, "allwinner,sun6i-a31-ccu",
+		      sun6i_a31_ccu_setup);
+
+/*
+ * Regmap for the GMAC driver (dwmac-sunxi) to allow access to
+ * GMAC configuration register.
+ */
+#define SUN6I_A31_GMAC_CFG_REG 0xD0
+static bool sun6i_a31_ccu_regmap_accessible_reg(struct device *dev,
+						unsigned int reg)
+{
+	if (reg == SUN6I_A31_GMAC_CFG_REG)
+		return true;
+	return false;
+}
+
+static struct regmap_config sun6i_a31_ccu_regmap_config = {
+	.reg_bits	= 32,
+	.val_bits	= 32,
+	.reg_stride	= 4,
+	.max_register	= 0x308, /* clk_out_b */
+
+	.readable_reg	= sun6i_a31_ccu_regmap_accessible_reg,
+	.writeable_reg	= sun6i_a31_ccu_regmap_accessible_reg,
+};
+
+static int sun6i_a31_ccu_probe_regmap(struct platform_device *pdev)
+{
+	void __iomem *reg;
+	struct resource *res;
+	struct regmap *regmap;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	reg = devm_ioremap(&pdev->dev, res->start, resource_size(res));
+	if (IS_ERR(reg))
+		return PTR_ERR(reg);
+
+	regmap = devm_regmap_init_mmio(&pdev->dev, reg,
+				       &sun6i_a31_ccu_regmap_config);
+	if (IS_ERR(regmap))
+		return PTR_ERR(regmap);
+
+	return 0;
+}
+
+static const struct of_device_id sun6i_a31_ccu_ids[] = {
+	{ .compatible = "allwinner,sun6i-a31-ccu"},
+	{ }
+};
+
+static struct platform_driver sun6i_a31_ccu_driver = {
+	.probe = sun6i_a31_ccu_probe_regmap,
+	.driver = {
+		.name = "sun6i-a31-ccu",
+		.of_match_table = sun6i_a31_ccu_ids,
+	},
+};
+builtin_platform_driver(sun6i_a31_ccu_driver);
-- 
2.26.2

[PATCH v3 3/6] net: stmmac: dwmac-sunxi: Implement syscon-based clock handling

[Priit Laes]

Convert the sun7i-gmac driver to use a regmap-based driver,
instead of relying on the custom clock implementation.

This allows to get rid of the last custom clock in the sun7i
device tree making the sun7i fully CCU-compatible.

Compatibility with existing devicetrees is retained.

Signed-off-by: Priit Laes <plaes@plaes.org>
---
 .../net/ethernet/stmicro/stmmac/dwmac-sunxi.c | 130 ++++++++++++++++--
 1 file changed, 122 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c
index 0e1ca2cba3c7..206398f7a2af 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c
@@ -12,8 +12,11 @@
 #include <linux/module.h>
 #include <linux/phy.h>
 #include <linux/platform_device.h>
+#include <linux/of_device.h>
 #include <linux/of_net.h>
 #include <linux/regulator/consumer.h>
+#include <linux/regmap.h>
+#include <linux/mfd/syscon.h>
 
 #include "stmmac_platform.h"
 
@@ -22,11 +25,23 @@ struct sunxi_priv_data {
 	int clk_enabled;
 	struct clk *tx_clk;
 	struct regulator *regulator;
+	struct regmap_field *regmap_field;
+};
+
+/* EMAC clock register @ 0x164 in the CCU address range */
+static const struct reg_field ccu_reg_field = {
+	.reg = 0x164,
+	.lsb = 0,
+	.msb = 31,
 };
 
 #define SUN7I_GMAC_GMII_RGMII_RATE	125000000
 #define SUN7I_GMAC_MII_RATE		25000000
 
+#define SUN7I_A20_CLK_MASK		GENMASK(2, 0)
+#define SUN7I_A20_RGMII_CLK		(BIT(2) | BIT(1))
+#define SUN7I_A20_MII_CLK		0
+
 static int sun7i_gmac_init(struct platform_device *pdev, void *priv)
 {
 	struct sunxi_priv_data *gmac = priv;
@@ -38,7 +53,20 @@ static int sun7i_gmac_init(struct platform_device *pdev, void *priv)
 			return ret;
 	}
 
-	/* Set GMAC interface port mode
+	if (gmac->regmap_field) {
+		if (phy_interface_mode_is_rgmii(gmac->interface)) {
+			regmap_field_update_bits(gmac->regmap_field,
+						 SUN7I_A20_CLK_MASK,
+						 SUN7I_A20_RGMII_CLK);
+			return clk_prepare_enable(gmac->tx_clk);
+		}
+		regmap_field_update_bits(gmac->regmap_field,
+					 SUN7I_A20_CLK_MASK,
+					 SUN7I_A20_MII_CLK);
+		return clk_enable(gmac->tx_clk);
+	}
+
+	/* Legacy devicetree clock (allwinner,sun7i-a20-gmac-clk) support:
 	 *
 	 * The GMAC TX clock lines are configured by setting the clock
 	 * rate, which then uses the auto-reparenting feature of the
@@ -62,9 +90,16 @@ static void sun7i_gmac_exit(struct platform_device *pdev, void *priv)
 {
 	struct sunxi_priv_data *gmac = priv;
 
-	if (gmac->clk_enabled) {
+	if (gmac->regmap_field) {
+		regmap_field_update_bits(gmac->regmap_field,
+					 SUN7I_A20_CLK_MASK, 0);
 		clk_disable(gmac->tx_clk);
-		gmac->clk_enabled = 0;
+	} else {
+		/* Handle legacy devicetree clock (sun7i-a20-gmac-clk) */
+		if (gmac->clk_enabled) {
+			clk_disable(gmac->tx_clk);
+			gmac->clk_enabled = 0;
+		}
 	}
 	clk_unprepare(gmac->tx_clk);
 
@@ -72,10 +107,55 @@ static void sun7i_gmac_exit(struct platform_device *pdev, void *priv)
 		regulator_disable(gmac->regulator);
 }
 
+static struct regmap *sun7i_gmac_get_syscon_from_dev(struct device_node *node)
+{
+	struct device_node *syscon_node;
+	struct platform_device *syscon_pdev;
+	struct regmap *regmap = NULL;
+
+	syscon_node = of_parse_phandle(node, "syscon", 0);
+	if (!syscon_node)
+		return ERR_PTR(-ENODEV);
+
+	syscon_pdev = of_find_device_by_node(syscon_node);
+	if (!syscon_pdev) {
+		/* platform device might not be probed yet */
+		regmap = ERR_PTR(-EPROBE_DEFER);
+		goto out_put_node;
+	}
+
+	/* If no regmap is found then the other device driver is at fault */
+	regmap = dev_get_regmap(&syscon_pdev->dev, NULL);
+	if (!regmap)
+		regmap = ERR_PTR(-EINVAL);
+
+	platform_device_put(syscon_pdev);
+out_put_node:
+	of_node_put(syscon_node);
+	return regmap;
+}
+
 static void sun7i_fix_speed(void *priv, unsigned int speed)
 {
 	struct sunxi_priv_data *gmac = priv;
 
+	if (gmac->regmap_field) {
+		clk_disable(gmac->tx_clk);
+		clk_unprepare(gmac->tx_clk);
+		if (speed == 1000)
+			regmap_field_update_bits(gmac->regmap_field,
+						 SUN7I_A20_CLK_MASK,
+						 SUN7I_A20_RGMII_CLK);
+		else
+			regmap_field_update_bits(gmac->regmap_field,
+						 SUN7I_A20_CLK_MASK,
+						 SUN7I_A20_MII_CLK);
+		clk_prepare_enable(gmac->tx_clk);
+		return;
+	}
+
+	/* Handle legacy devicetree clock (sun7i-a20-gmac-clk) */
+
 	/* only GMII mode requires us to reconfigure the clock lines */
 	if (gmac->interface != PHY_INTERFACE_MODE_GMII)
 		return;
@@ -102,6 +182,8 @@ static int sun7i_gmac_probe(struct platform_device *pdev)
 	struct stmmac_resources stmmac_res;
 	struct sunxi_priv_data *gmac;
 	struct device *dev = &pdev->dev;
+	struct device_node *syscon_node;
+	struct regmap *regmap = NULL;
 	int ret;
 
 	ret = stmmac_get_platform_resources(pdev, &stmmac_res);
@@ -124,11 +206,43 @@ static int sun7i_gmac_probe(struct platform_device *pdev)
 		goto err_remove_config_dt;
 	}
 
-	gmac->tx_clk = devm_clk_get(dev, "allwinner_gmac_tx");
-	if (IS_ERR(gmac->tx_clk)) {
-		dev_err(dev, "could not get tx clock\n");
-		ret = PTR_ERR(gmac->tx_clk);
-		goto err_remove_config_dt;
+	/* Attempt to fetch syscon node... */
+	syscon_node = of_parse_phandle(dev->of_node, "syscon", 0);
+	if (syscon_node) {
+		gmac->tx_clk = devm_clk_get(dev, "stmmaceth");
+		if (IS_ERR(gmac->tx_clk)) {
+			dev_err(dev, "Could not get TX clock\n");
+			ret = PTR_ERR(gmac->tx_clk);
+			goto err_remove_config_dt;
+		}
+
+		regmap = sun7i_gmac_get_syscon_from_dev(pdev->dev.of_node);
+		if (IS_ERR(regmap))
+			regmap = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+								 "syscon");
+		if (IS_ERR(regmap)) {
+			ret = PTR_ERR(regmap);
+			dev_err(&pdev->dev, "Unable to map syscon: %d\n", ret);
+			goto err_remove_config_dt;
+		}
+
+		gmac->regmap_field = devm_regmap_field_alloc(dev, regmap,
+							     ccu_reg_field);
+
+		if (IS_ERR(gmac->regmap_field)) {
+			ret = PTR_ERR(gmac->regmap_field);
+			dev_err(dev, "Unable to map syscon register: %d\n", ret);
+			goto err_remove_config_dt;
+		}
+	/* ...or fall back to legacy clock setup */
+	} else {
+		gmac->tx_clk = devm_clk_get(dev, "allwinner_gmac_tx");
+		if (IS_ERR(gmac->tx_clk)) {
+			dev_err(dev, "could not get tx clock\n");
+			ret = PTR_ERR(gmac->tx_clk);
+			goto err_remove_config_dt;
+		}
+		dev_info(dev, "allwinner_gmac_tx support is deprecated!\n");
 	}
 
 	/* Optional regulator for PHY */
-- 
2.26.2

Re: [PATCH v3 3/6] net: stmmac: dwmac-sunxi: Implement syscon-based clock handling

[Maxime Ripard]

[-- Attachment #1: Type: text/plain, Size: 6420 bytes --]

On Thu, Apr 30, 2020 at 02:56:59PM +0300, Priit Laes wrote:
> Convert the sun7i-gmac driver to use a regmap-based driver,
> instead of relying on the custom clock implementation.
> 
> This allows to get rid of the last custom clock in the sun7i
> device tree making the sun7i fully CCU-compatible.
> 
> Compatibility with existing devicetrees is retained.
> 
> Signed-off-by: Priit Laes <plaes@plaes.org>
> ---
>  .../net/ethernet/stmicro/stmmac/dwmac-sunxi.c | 130 ++++++++++++++++--
>  1 file changed, 122 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c
> index 0e1ca2cba3c7..206398f7a2af 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-sunxi.c
> @@ -12,8 +12,11 @@
>  #include <linux/module.h>
>  #include <linux/phy.h>
>  #include <linux/platform_device.h>
> +#include <linux/of_device.h>
>  #include <linux/of_net.h>
>  #include <linux/regulator/consumer.h>
> +#include <linux/regmap.h>
> +#include <linux/mfd/syscon.h>
>  
>  #include "stmmac_platform.h"
>  
> @@ -22,11 +25,23 @@ struct sunxi_priv_data {
>  	int clk_enabled;
>  	struct clk *tx_clk;
>  	struct regulator *regulator;
> +	struct regmap_field *regmap_field;
> +};
> +
> +/* EMAC clock register @ 0x164 in the CCU address range */
> +static const struct reg_field ccu_reg_field = {
> +	.reg = 0x164,
> +	.lsb = 0,
> +	.msb = 31,
>  };
>  
>  #define SUN7I_GMAC_GMII_RGMII_RATE	125000000
>  #define SUN7I_GMAC_MII_RATE		25000000
>  
> +#define SUN7I_A20_CLK_MASK		GENMASK(2, 0)
> +#define SUN7I_A20_RGMII_CLK		(BIT(2) | BIT(1))
> +#define SUN7I_A20_MII_CLK		0
> +
>  static int sun7i_gmac_init(struct platform_device *pdev, void *priv)
>  {
>  	struct sunxi_priv_data *gmac = priv;
> @@ -38,7 +53,20 @@ static int sun7i_gmac_init(struct platform_device *pdev, void *priv)
>  			return ret;
>  	}
>  
> -	/* Set GMAC interface port mode
> +	if (gmac->regmap_field) {
> +		if (phy_interface_mode_is_rgmii(gmac->interface)) {
> +			regmap_field_update_bits(gmac->regmap_field,
> +						 SUN7I_A20_CLK_MASK,
> +						 SUN7I_A20_RGMII_CLK);
> +			return clk_prepare_enable(gmac->tx_clk);
> +		}

Why do you prepare and enable the clock here? ...

> +		regmap_field_update_bits(gmac->regmap_field,
> +					 SUN7I_A20_CLK_MASK,
> +					 SUN7I_A20_MII_CLK);
> +		return clk_enable(gmac->tx_clk);
> +	}

... while you only enable it here ...

> +	/* Legacy devicetree clock (allwinner,sun7i-a20-gmac-clk) support:
>  	 *
>  	 * The GMAC TX clock lines are configured by setting the clock
>  	 * rate, which then uses the auto-reparenting feature of the
> @@ -62,9 +90,16 @@ static void sun7i_gmac_exit(struct platform_device *pdev, void *priv)
>  {
>  	struct sunxi_priv_data *gmac = priv;
>  
> -	if (gmac->clk_enabled) {
> +	if (gmac->regmap_field) {
> +		regmap_field_update_bits(gmac->regmap_field,
> +					 SUN7I_A20_CLK_MASK, 0);
>  		clk_disable(gmac->tx_clk);
> -		gmac->clk_enabled = 0;
> +	} else {
> +		/* Handle legacy devicetree clock (sun7i-a20-gmac-clk) */
> +		if (gmac->clk_enabled) {
> +			clk_disable(gmac->tx_clk);
> +			gmac->clk_enabled = 0;
> +		}
>  	}
>  	clk_unprepare(gmac->tx_clk);

.... and disable and unprepare it here?

> @@ -72,10 +107,55 @@ static void sun7i_gmac_exit(struct platform_device *pdev, void *priv)
>  		regulator_disable(gmac->regulator);
>  }
>  
> +static struct regmap *sun7i_gmac_get_syscon_from_dev(struct device_node *node)
> +{
> +	struct device_node *syscon_node;
> +	struct platform_device *syscon_pdev;
> +	struct regmap *regmap = NULL;
> +
> +	syscon_node = of_parse_phandle(node, "syscon", 0);
> +	if (!syscon_node)
> +		return ERR_PTR(-ENODEV);
> +
> +	syscon_pdev = of_find_device_by_node(syscon_node);
> +	if (!syscon_pdev) {
> +		/* platform device might not be probed yet */
> +		regmap = ERR_PTR(-EPROBE_DEFER);
> +		goto out_put_node;
> +	}
> +
> +	/* If no regmap is found then the other device driver is at fault */
> +	regmap = dev_get_regmap(&syscon_pdev->dev, NULL);
> +	if (!regmap)
> +		regmap = ERR_PTR(-EINVAL);
> +
> +	platform_device_put(syscon_pdev);
> +out_put_node:
> +	of_node_put(syscon_node);
> +	return regmap;
> +}
> +
>  static void sun7i_fix_speed(void *priv, unsigned int speed)
>  {
>  	struct sunxi_priv_data *gmac = priv;
>  
> +	if (gmac->regmap_field) {
> +		clk_disable(gmac->tx_clk);
> +		clk_unprepare(gmac->tx_clk);
> +		if (speed == 1000)
> +			regmap_field_update_bits(gmac->regmap_field,
> +						 SUN7I_A20_CLK_MASK,
> +						 SUN7I_A20_RGMII_CLK);
> +		else
> +			regmap_field_update_bits(gmac->regmap_field,
> +						 SUN7I_A20_CLK_MASK,
> +						 SUN7I_A20_MII_CLK);
> +		clk_prepare_enable(gmac->tx_clk);


If were going to use clk_prepare_enable, we might as well use
clk_disable_unprepare

> +		return;
> +	}
> +
> +	/* Handle legacy devicetree clock (sun7i-a20-gmac-clk) */

That doesn't say much, you should rather explain what the situation is exactly.

> +
>  	/* only GMII mode requires us to reconfigure the clock lines */
>  	if (gmac->interface != PHY_INTERFACE_MODE_GMII)
>  		return;
> @@ -102,6 +182,8 @@ static int sun7i_gmac_probe(struct platform_device *pdev)
>  	struct stmmac_resources stmmac_res;
>  	struct sunxi_priv_data *gmac;
>  	struct device *dev = &pdev->dev;
> +	struct device_node *syscon_node;
> +	struct regmap *regmap = NULL;
>  	int ret;
>  
>  	ret = stmmac_get_platform_resources(pdev, &stmmac_res);
> @@ -124,11 +206,43 @@ static int sun7i_gmac_probe(struct platform_device *pdev)
>  		goto err_remove_config_dt;
>  	}
>  
> -	gmac->tx_clk = devm_clk_get(dev, "allwinner_gmac_tx");
> -	if (IS_ERR(gmac->tx_clk)) {
> -		dev_err(dev, "could not get tx clock\n");
> -		ret = PTR_ERR(gmac->tx_clk);
> -		goto err_remove_config_dt;
> +	/* Attempt to fetch syscon node... */
> +	syscon_node = of_parse_phandle(dev->of_node, "syscon", 0);
> +	if (syscon_node) {
> +		gmac->tx_clk = devm_clk_get(dev, "stmmaceth");
> +		if (IS_ERR(gmac->tx_clk)) {
> +			dev_err(dev, "Could not get TX clock\n");
> +			ret = PTR_ERR(gmac->tx_clk);
> +			goto err_remove_config_dt;
> +		}

I'm not quite sure why you added this clock lookup here? Wasn't it here already?

Maxime

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

[PATCH v3 4/6] dt-bindings: net: sun7i-gmac: Add syscon support

[Priit Laes]

Now that driver supports syscon-based regmap access, document also the
devicetree binding.

Signed-off-by: Priit Laes <plaes@plaes.org>
---
 .../bindings/net/allwinner,sun7i-a20-gmac.yaml    | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml b/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml
index 703d0d886884..c41d7c598c19 100644
--- a/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml
+++ b/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml
@@ -29,17 +29,26 @@ properties:
   clocks:
     items:
       - description: GMAC main clock
+
+      # Deprecated
       - description: TX clock
 
   clock-names:
     items:
       - const: stmmaceth
+
+      # Deprecated
       - const: allwinner_gmac_tx
 
   phy-supply:
     description:
       PHY regulator
 
+  syscon:
+    $ref: /schemas/types.yaml#definitions/phandle
+    description:
+      Phandle to the device containing the GMAC clock register
+
 required:
   - compatible
   - reg
@@ -48,6 +57,7 @@ required:
   - clocks
   - clock-names
   - phy-mode
+  - syscon
 
 unevaluatedProperties: false
 
@@ -55,11 +65,12 @@ examples:
   - |
     gmac: ethernet@1c50000 {
         compatible = "allwinner,sun7i-a20-gmac";
+        syscon = <&syscon>;
         reg = <0x01c50000 0x10000>;
         interrupts = <0 85 1>;
         interrupt-names = "macirq";
-        clocks = <&ahb_gates 49>, <&gmac_tx>;
-        clock-names = "stmmaceth", "allwinner_gmac_tx";
+        clocks = <&ahb_gates 49>;
+        clock-names = "stmmaceth";
         phy-mode = "mii";
     };
 
-- 
2.26.2

Re: [PATCH v3 4/6] dt-bindings: net: sun7i-gmac: Add syscon support

[Maxime Ripard]

[-- Attachment #1: Type: text/plain, Size: 1645 bytes --]

On Thu, Apr 30, 2020 at 02:57:00PM +0300, Priit Laes wrote:
> Now that driver supports syscon-based regmap access, document also the
> devicetree binding.
> 
> Signed-off-by: Priit Laes <plaes@plaes.org>
> ---
>  .../bindings/net/allwinner,sun7i-a20-gmac.yaml    | 15 +++++++++++++--
>  1 file changed, 13 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml b/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml
> index 703d0d886884..c41d7c598c19 100644
> --- a/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml
> +++ b/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.yaml
> @@ -29,17 +29,26 @@ properties:
>    clocks:
>      items:
>        - description: GMAC main clock
> +
> +      # Deprecated
>        - description: TX clock
>  
>    clock-names:
>      items:
>        - const: stmmaceth
> +
> +      # Deprecated
>        - const: allwinner_gmac_tx

That would be deprecated: true

Also, you still require to have two clocks here?

>  
>    phy-supply:
>      description:
>        PHY regulator
>  
> +  syscon:
> +    $ref: /schemas/types.yaml#definitions/phandle
> +    description:
> +      Phandle to the device containing the GMAC clock register
> +
>  required:
>    - compatible
>    - reg
> @@ -48,6 +57,7 @@ required:
>    - clocks
>    - clock-names
>    - phy-mode
> +  - syscon

You can't do that though, this changes the binding in a non-backward compatible
way

I guess you could add an if clause to require it if clocks has a single member.

Maxime

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH v3 4/6] dt-bindings: net: sun7i-gmac: Add syscon support

[Rob Herring]

On Thu, 30 Apr 2020 14:57:00 +0300, Priit Laes wrote:
> Now that driver supports syscon-based regmap access, document also the
> devicetree binding.
> 
> Signed-off-by: Priit Laes <plaes@plaes.org>
> ---
>  .../bindings/net/allwinner,sun7i-a20-gmac.yaml    | 15 +++++++++++++--
>  1 file changed, 13 insertions(+), 2 deletions(-)
> 

My bot found errors running 'make dt_binding_check' on your patch:

/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.example.dt.yaml: ethernet@1c50000: clock-names: ['stmmaceth'] is too short
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/net/allwinner,sun7i-a20-gmac.example.dt.yaml: ethernet@1c50000: clocks: [[4294967295, 49]] is too short

See https://patchwork.ozlabs.org/patch/1280292

If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure dt-schema is up to date:

pip3 install git+https://github.com/devicetree-org/dt-schema.git@master --upgrade

Please check and re-submit.

[PATCH v3 5/6] ARM: dts: sun7i: Use syscon-based implementation for gmac

[Priit Laes]

Use syscon-based approach to access gmac clock configuration
register, instead of relying on a custom clock driver.

As a bonus, we can now drop the custom clock implementation
and dummy clocks making sun7i fully CCU-compatible.

Signed-off-by: Priit Laes <plaes@plaes.org>
---
 arch/arm/boot/dts/sun7i-a20.dtsi | 36 +++-----------------------------
 1 file changed, 3 insertions(+), 33 deletions(-)

diff --git a/arch/arm/boot/dts/sun7i-a20.dtsi b/arch/arm/boot/dts/sun7i-a20.dtsi
index ffe1d10a1a84..750962a94fad 100644
--- a/arch/arm/boot/dts/sun7i-a20.dtsi
+++ b/arch/arm/boot/dts/sun7i-a20.dtsi
@@ -219,37 +219,6 @@ osc32k: clk-32k {
 			clock-frequency = <32768>;
 			clock-output-names = "osc32k";
 		};
-
-		/*
-		 * The following two are dummy clocks, placeholders
-		 * used in the gmac_tx clock. The gmac driver will
-		 * choose one parent depending on the PHY interface
-		 * mode, using clk_set_rate auto-reparenting.
-		 *
-		 * The actual TX clock rate is not controlled by the
-		 * gmac_tx clock.
-		 */
-		mii_phy_tx_clk: clk-mii-phy-tx {
-			#clock-cells = <0>;
-			compatible = "fixed-clock";
-			clock-frequency = <25000000>;
-			clock-output-names = "mii_phy_tx";
-		};
-
-		gmac_int_tx_clk: clk-gmac-int-tx {
-			#clock-cells = <0>;
-			compatible = "fixed-clock";
-			clock-frequency = <125000000>;
-			clock-output-names = "gmac_int_tx";
-		};
-
-		gmac_tx_clk: clk@1c20164 {
-			#clock-cells = <0>;
-			compatible = "allwinner,sun7i-a20-gmac-clk";
-			reg = <0x01c20164 0x4>;
-			clocks = <&mii_phy_tx_clk>, <&gmac_int_tx_clk>;
-			clock-output-names = "gmac_tx";
-		};
 	};
 
 
@@ -1511,11 +1480,12 @@ mali: gpu@1c40000 {
 
 		gmac: ethernet@1c50000 {
 			compatible = "allwinner,sun7i-a20-gmac";
+			syscon = <&ccu>;
 			reg = <0x01c50000 0x10000>;
 			interrupts = <GIC_SPI 85 IRQ_TYPE_LEVEL_HIGH>;
 			interrupt-names = "macirq";
-			clocks = <&ccu CLK_AHB_GMAC>, <&gmac_tx_clk>;
-			clock-names = "stmmaceth", "allwinner_gmac_tx";
+			clocks = <&ccu CLK_AHB_GMAC>;
+			clock-names = "stmmaceth";
 			snps,pbl = <2>;
 			snps,fixed-burst;
 			snps,force_sf_dma_mode;
-- 
2.26.2

[PATCH v3 6/6] ARM: dts: sun6i: Use syscon-based implementation for gmac

[Priit Laes]

Use syscon-based approach to access gmac clock configuration
register instead of relying on a custom clock driver.

As a bonus, we can now drop the custom clock implementation
and the dummy clocks.

Signed-off-by: Priit Laes <plaes@plaes.org>
---
 arch/arm/boot/dts/sun6i-a31.dtsi | 35 +++-----------------------------
 1 file changed, 3 insertions(+), 32 deletions(-)

diff --git a/arch/arm/boot/dts/sun6i-a31.dtsi b/arch/arm/boot/dts/sun6i-a31.dtsi
index f3425a66fc0a..fcf8a242741f 100644
--- a/arch/arm/boot/dts/sun6i-a31.dtsi
+++ b/arch/arm/boot/dts/sun6i-a31.dtsi
@@ -228,36 +228,6 @@ osc32k: clk-32k {
 			clock-output-names = "ext_osc32k";
 		};
 
-		/*
-		 * The following two are dummy clocks, placeholders
-		 * used in the gmac_tx clock. The gmac driver will
-		 * choose one parent depending on the PHY interface
-		 * mode, using clk_set_rate auto-reparenting.
-		 *
-		 * The actual TX clock rate is not controlled by the
-		 * gmac_tx clock.
-		 */
-		mii_phy_tx_clk: clk-mii-phy-tx {
-			#clock-cells = <0>;
-			compatible = "fixed-clock";
-			clock-frequency = <25000000>;
-			clock-output-names = "mii_phy_tx";
-		};
-
-		gmac_int_tx_clk: clk-gmac-int-tx {
-			#clock-cells = <0>;
-			compatible = "fixed-clock";
-			clock-frequency = <125000000>;
-			clock-output-names = "gmac_int_tx";
-		};
-
-		gmac_tx_clk: clk@1c200d0 {
-			#clock-cells = <0>;
-			compatible = "allwinner,sun7i-a20-gmac-clk";
-			reg = <0x01c200d0 0x4>;
-			clocks = <&mii_phy_tx_clk>, <&gmac_int_tx_clk>;
-			clock-output-names = "gmac_tx";
-		};
 	};
 
 	de: display-engine {
@@ -943,11 +913,12 @@ i2c3: i2c@1c2b800 {
 
 		gmac: ethernet@1c30000 {
 			compatible = "allwinner,sun7i-a20-gmac";
+			syscon = <&ccu>;
 			reg = <0x01c30000 0x1054>;
 			interrupts = <GIC_SPI 82 IRQ_TYPE_LEVEL_HIGH>;
 			interrupt-names = "macirq";
-			clocks = <&ccu CLK_AHB1_EMAC>, <&gmac_tx_clk>;
-			clock-names = "stmmaceth", "allwinner_gmac_tx";
+			clocks = <&ccu CLK_AHB1_EMAC>;
+			clock-names = "stmmaceth";
 			resets = <&ccu RST_AHB1_EMAC>;
 			reset-names = "stmmaceth";
 			snps,pbl = <2>;
-- 
2.26.2