[PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[Linus Walleij]

This adds device tree bindings for the IXP4xx ethernet
controller with optional MDIO bridge.

Cc: Zoltan HERPAI <wigyori@uid0.hu>
Cc: Raylynn Knight <rayknight@me.com>
Cc: devicetree@vger.kernel.org
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
---
ChangeLog v1->v2:
- Add schema for the (optional) embedded MDIO bus inside
  the ethernet controller in an "mdio" node instead of just
  letting the code randomly populate and present it to
  the operating system.
- Reference the standard schemas for ethernet controller and
  MDIO buses.
- Add intel,npe to indentify the NPE unit used with each
  ethernet adapter.
---
 .../bindings/net/intel,ixp4xx-ethernet.yaml   | 80 +++++++++++++++++++
 1 file changed, 80 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.yaml

diff --git a/Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.yaml b/Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.yaml
new file mode 100644
index 000000000000..55ef6ff7d171
--- /dev/null
+++ b/Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.yaml
@@ -0,0 +1,80 @@
+# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
+# Copyright 2018 Linaro Ltd.
+%YAML 1.2
+---
+$id: "http://devicetree.org/schemas/net/intel,ixp4xx-ethernet.yaml#"
+$schema: "http://devicetree.org/meta-schemas/core.yaml#"
+
+title: Intel IXP4xx ethernet
+
+allOf:
+  - $ref: "ethernet-controller.yaml#"
+
+maintainers:
+  - Linus Walleij <linus.walleij@linaro.org>
+
+description: |
+  The Intel IXP4xx ethernet makes use of the IXP4xx NPE (Network
+  Processing Engine) and the IXP4xx Queue Mangager to process
+  the ethernet frames. It can optionally contain an MDIO bus to
+  talk to PHYs.
+
+properties:
+  compatible:
+    const: intel,ixp4xx-ethernet
+
+  reg:
+    maxItems: 1
+    description: Ethernet MMIO address range
+
+  queue-rx:
+    $ref: '/schemas/types.yaml#/definitions/phandle-array'
+    maxItems: 1
+    description: phandle to the RX queue on the NPE
+
+  queue-txready:
+    $ref: '/schemas/types.yaml#/definitions/phandle-array'
+    maxItems: 1
+    description: phandle to the TX READY queue on the NPE
+
+  phy-handle: true
+
+  intel,npe:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    enum: [0, 1, 2, 3]
+    description: which NPE (Network Processing Engine) this ethernet
+      instance is using
+
+  mdio:
+    type: object
+    $ref: "mdio.yaml#"
+    description: optional node for embedded MDIO controller
+
+required:
+  - compatible
+  - reg
+  - queue-rx
+  - queue-txready
+
+additionalProperties: false
+
+examples:
+  - |
+    ethernet@c8009000 {
+      compatible = "intel,ixp4xx-ethernet";
+      reg = <0xc8009000 0x1000>;
+      status = "disabled";
+      queue-rx = <&qmgr 3>;
+      queue-txready = <&qmgr 20>;
+      intel,npe = <1>;
+      phy-handle = <&phy1>;
+
+      mdio {
+        #address-cells = <1>;
+        #size-cells = <0>;
+        phy1: phy@1 {
+          #phy-cells = <0>;
+          reg = <1>;
+        };
+      };
+    };
-- 
2.29.2

[PATCH 2/3] net: ethernet: ixp4xx: Support device tree probing

[Linus Walleij]

This adds device tree probing to the IXP4xx ethernet
driver.

Add a platform data bool to tell us whether to
register an MDIO bus for the device or not, as well
as the corresponding NPE.

We need to drop the memory region request as part of
this since the OF core will request the memory for the
device.

Cc: Zoltan HERPAI <wigyori@uid0.hu>
Cc: Raylynn Knight <rayknight@me.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
---
ChangeLog v1->v2:
- Do not rely on earlier attempt for identifying which
  ports to use: we split the support clearly in a DT way
  and a playform data way.
- Isolate the DT code better.
---
 drivers/net/ethernet/xscale/ixp4xx_eth.c | 202 ++++++++++++++++-------
 include/linux/platform_data/eth_ixp4xx.h |   2 +
 2 files changed, 148 insertions(+), 56 deletions(-)

diff --git a/drivers/net/ethernet/xscale/ixp4xx_eth.c b/drivers/net/ethernet/xscale/ixp4xx_eth.c
index 9defaa21a1a9..758f820068b5 100644
--- a/drivers/net/ethernet/xscale/ixp4xx_eth.c
+++ b/drivers/net/ethernet/xscale/ixp4xx_eth.c
@@ -28,6 +28,7 @@
 #include <linux/kernel.h>
 #include <linux/net_tstamp.h>
 #include <linux/of.h>
+#include <linux/of_mdio.h>
 #include <linux/phy.h>
 #include <linux/platform_data/eth_ixp4xx.h>
 #include <linux/platform_device.h>
@@ -165,7 +166,6 @@ struct eth_regs {
 };
 
 struct port {
-	struct resource *mem_res;
 	struct eth_regs __iomem *regs;
 	struct npe *npe;
 	struct net_device *netdev;
@@ -1358,19 +1358,132 @@ static const struct net_device_ops ixp4xx_netdev_ops = {
 	.ndo_validate_addr = eth_validate_addr,
 };
 
+#ifdef CONFIG_OF
+static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
+{
+	struct device_node *np = dev->of_node;
+	struct of_phandle_args queue_spec;
+	struct eth_plat_info *plat;
+	struct device_node *phy_np;
+	struct device_node *mdio_np;
+	u32 val;
+	int ret;
+
+	plat = devm_kzalloc(dev, sizeof(*plat), GFP_KERNEL);
+	if (!plat)
+		return NULL;
+
+	ret = of_property_read_u32(np, "intel,npe", &val);
+	if (ret) {
+		dev_err(dev, "no NPE engine specified\n");
+		return NULL;
+	}
+	/* NPE ID 0x00, 0x10, 0x20... */
+	plat->npe = (val << 4);
+
+	phy_np = of_parse_phandle(np, "phy-handle", 0);
+	if (phy_np) {
+		ret = of_property_read_u32(phy_np, "reg", &val);
+		if (ret) {
+			dev_err(dev, "cannot find phy reg\n");
+			return NULL;
+		}
+		of_node_put(phy_np);
+	} else {
+		dev_err(dev, "cannot find phy instance\n");
+		val = 0;
+	}
+	plat->phy = val;
+
+	/* Check if this device has an MDIO bus */
+	mdio_np = of_get_child_by_name(np, "mdio");
+	if (mdio_np) {
+		plat->has_mdio = true;
+		of_node_put(mdio_np);
+	}
+
+	/* Get the rx queue as a resource from queue manager */
+	ret = of_parse_phandle_with_fixed_args(np, "queue-rx", 1, 0,
+					       &queue_spec);
+	if (ret) {
+		dev_err(dev, "no rx queue phandle\n");
+		return NULL;
+	}
+	plat->rxq = queue_spec.args[0];
+
+	/* Get the txready queue as resource from queue manager */
+	ret = of_parse_phandle_with_fixed_args(np, "queue-txready", 1, 0,
+					       &queue_spec);
+	if (ret) {
+		dev_err(dev, "no txready queue phandle\n");
+		return NULL;
+	}
+	plat->txreadyq = queue_spec.args[0];
+
+	return plat;
+}
+#else
+static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
+{
+	return NULL;
+}
+#endif
+
 static int ixp4xx_eth_probe(struct platform_device *pdev)
 {
 	char phy_id[MII_BUS_ID_SIZE + 3];
 	struct phy_device *phydev = NULL;
 	struct device *dev = &pdev->dev;
+	struct device_node *np = dev->of_node;
 	struct eth_plat_info *plat;
-	resource_size_t regs_phys;
 	struct net_device *ndev;
 	struct resource *res;
 	struct port *port;
 	int err;
 
-	plat = dev_get_platdata(dev);
+	if (np) {
+		plat = ixp4xx_of_get_platdata(dev);
+		if (!plat)
+			return -ENODEV;
+	} else {
+		plat = dev_get_platdata(dev);
+		if (!plat)
+			return -ENODEV;
+		plat->npe = pdev->id;
+		switch (plat->npe) {
+		case IXP4XX_ETH_NPEA:
+			/* If the MDIO bus is not up yet, defer probe */
+			break;
+		case IXP4XX_ETH_NPEB:
+			/* On all except IXP43x, NPE-B is used for the MDIO bus.
+			 * If there is no NPE-B in the feature set, bail out,
+			 * else we have the MDIO bus here.
+			 */
+			if (!cpu_is_ixp43x()) {
+				if (!(ixp4xx_read_feature_bits() &
+				      IXP4XX_FEATURE_NPEB_ETH0))
+					return -ENODEV;
+				/* Else register the MDIO bus on NPE-B */
+				plat->has_mdio = true;
+			}
+			break;
+		case IXP4XX_ETH_NPEC:
+			/* IXP43x lacks NPE-B and uses NPE-C for the MDIO bus
+			 * access, if there is no NPE-C, no bus, nothing works,
+			 * so bail out.
+			 */
+			if (cpu_is_ixp43x()) {
+				if (!(ixp4xx_read_feature_bits() &
+				      IXP4XX_FEATURE_NPEC_ETH))
+					return -ENODEV;
+				/* Else register the MDIO bus on NPE-B */
+				plat->has_mdio = true;
+			}
+			break;
+		default:
+			return -ENODEV;
+		}
+	}
 
 	if (!(ndev = devm_alloc_etherdev(dev, sizeof(struct port))))
 		return -ENOMEM;
@@ -1378,59 +1491,29 @@ static int ixp4xx_eth_probe(struct platform_device *pdev)
 	SET_NETDEV_DEV(ndev, dev);
 	port = netdev_priv(ndev);
 	port->netdev = ndev;
-	port->id = pdev->id;
+	port->id = plat->npe;
 
 	/* Get the port resource and remap */
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	if (!res)
 		return -ENODEV;
-	regs_phys = res->start;
 	port->regs = devm_ioremap_resource(dev, res);
 	if (IS_ERR(port->regs))
 		return PTR_ERR(port->regs);
 
-	switch (port->id) {
-	case IXP4XX_ETH_NPEA:
-		/* If the MDIO bus is not up yet, defer probe */
-		if (!mdio_bus)
-			return -EPROBE_DEFER;
-		break;
-	case IXP4XX_ETH_NPEB:
-		/*
-		 * On all except IXP43x, NPE-B is used for the MDIO bus.
-		 * If there is no NPE-B in the feature set, bail out, else
-		 * register the MDIO bus.
-		 */
-		if (!cpu_is_ixp43x()) {
-			if (!(ixp4xx_read_feature_bits() &
-			      IXP4XX_FEATURE_NPEB_ETH0))
-				return -ENODEV;
-			/* Else register the MDIO bus on NPE-B */
-			if ((err = ixp4xx_mdio_register(port->regs)))
-				return err;
-		}
-		if (!mdio_bus)
-			return -EPROBE_DEFER;
-		break;
-	case IXP4XX_ETH_NPEC:
-		/*
-		 * IXP43x lacks NPE-B and uses NPE-C for the MDIO bus access,
-		 * of there is no NPE-C, no bus, nothing works, so bail out.
-		 */
-		if (cpu_is_ixp43x()) {
-			if (!(ixp4xx_read_feature_bits() &
-			      IXP4XX_FEATURE_NPEC_ETH))
-				return -ENODEV;
-			/* Else register the MDIO bus on NPE-C */
-			if ((err = ixp4xx_mdio_register(port->regs)))
-				return err;
+	/* Register the MDIO bus if we have it */
+	if (plat->has_mdio) {
+		err = ixp4xx_mdio_register(port->regs);
+		if (err) {
+			dev_err(dev, "failed to register MDIO bus\n");
+			return err;
 		}
-		if (!mdio_bus)
-			return -EPROBE_DEFER;
-		break;
-	default:
-		return -ENODEV;
 	}
+	/* If the instance with the MDIO bus has not yet appeared,
+	 * defer probing until it gets probed.
+	 */
+	if (!mdio_bus)
+		return -EPROBE_DEFER;
 
 	ndev->netdev_ops = &ixp4xx_netdev_ops;
 	ndev->ethtool_ops = &ixp4xx_ethtool_ops;
@@ -1444,12 +1527,6 @@ static int ixp4xx_eth_probe(struct platform_device *pdev)
 	if (!(port->npe = npe_request(NPE_ID(port->id))))
 		return -EIO;
 
-	port->mem_res = request_mem_region(regs_phys, REGS_SIZE, ndev->name);
-	if (!port->mem_res) {
-		err = -EBUSY;
-		goto err_npe_rel;
-	}
-
 	port->plat = plat;
 	npe_port_tab[NPE_ID(port->id)] = port;
 	memcpy(ndev->dev_addr, plat->hwaddr, ETH_ALEN);
@@ -1463,11 +1540,17 @@ static int ixp4xx_eth_probe(struct platform_device *pdev)
 	udelay(50);
 
 	snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
-		mdio_bus->id, plat->phy);
+		 mdio_bus->id, plat->phy);
 	phydev = phy_connect(ndev, phy_id, &ixp4xx_adjust_link,
 			     PHY_INTERFACE_MODE_MII);
+	if (!phydev) {
+		err = -ENODEV;
+		dev_err(dev, "no phydev\n");
+		goto err_free_mem;
+	}
 	if (IS_ERR(phydev)) {
 		err = PTR_ERR(phydev);
+		dev_err(dev, "could not connect phydev (%d)\n", err);
 		goto err_free_mem;
 	}
 
@@ -1485,8 +1568,6 @@ static int ixp4xx_eth_probe(struct platform_device *pdev)
 	phy_disconnect(phydev);
 err_free_mem:
 	npe_port_tab[NPE_ID(port->id)] = NULL;
-	release_resource(port->mem_res);
-err_npe_rel:
 	npe_release(port->npe);
 	return err;
 }
@@ -1502,12 +1583,21 @@ static int ixp4xx_eth_remove(struct platform_device *pdev)
 	ixp4xx_mdio_remove();
 	npe_port_tab[NPE_ID(port->id)] = NULL;
 	npe_release(port->npe);
-	release_resource(port->mem_res);
 	return 0;
 }
 
+static const struct of_device_id ixp4xx_eth_of_match[] = {
+	{
+		.compatible = "intel,ixp4xx-ethernet",
+	},
+	{ },
+};
+
 static struct platform_driver ixp4xx_eth_driver = {
-	.driver.name	= DRV_NAME,
+	.driver = {
+		.name = DRV_NAME,
+		.of_match_table = of_match_ptr(ixp4xx_eth_of_match),
+	},
 	.probe		= ixp4xx_eth_probe,
 	.remove		= ixp4xx_eth_remove,
 };
diff --git a/include/linux/platform_data/eth_ixp4xx.h b/include/linux/platform_data/eth_ixp4xx.h
index 6f652ea0c6ae..114b0940729f 100644
--- a/include/linux/platform_data/eth_ixp4xx.h
+++ b/include/linux/platform_data/eth_ixp4xx.h
@@ -14,6 +14,8 @@ struct eth_plat_info {
 	u8 rxq;		/* configurable, currently 0 - 31 only */
 	u8 txreadyq;
 	u8 hwaddr[6];
+	u8 npe;		/* NPE instance used by this interface */
+	bool has_mdio;	/* If this instance has an MDIO bus */
 };
 
 #endif
-- 
2.29.2

[PATCH 3/3] net: ethernet: ixp4xx: Use OF MDIO bus registration

[Linus Walleij]

This augments the IXP4xx to use the OF MDIO bus code when
registering the MDIO bus and when looking up the PHY
for the ethernet network device.

Cc: Zoltan HERPAI <wigyori@uid0.hu>
Cc: Raylynn Knight <rayknight@me.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
---
ChangeLog v1->v2:
- New patch making use of more OF infrastructure.
---
 drivers/net/ethernet/xscale/Kconfig      |  1 +
 drivers/net/ethernet/xscale/ixp4xx_eth.c | 34 +++++++++---------------
 2 files changed, 14 insertions(+), 21 deletions(-)

diff --git a/drivers/net/ethernet/xscale/Kconfig b/drivers/net/ethernet/xscale/Kconfig
index 7b83a6e5d894..468ffe3d1707 100644
--- a/drivers/net/ethernet/xscale/Kconfig
+++ b/drivers/net/ethernet/xscale/Kconfig
@@ -22,6 +22,7 @@ config IXP4XX_ETH
 	tristate "Intel IXP4xx Ethernet support"
 	depends on ARM && ARCH_IXP4XX && IXP4XX_NPE && IXP4XX_QMGR
 	select PHYLIB
+	select OF_MDIO if OF
 	select NET_PTP_CLASSIFY
 	help
 	  Say Y here if you want to use built-in Ethernet ports
diff --git a/drivers/net/ethernet/xscale/ixp4xx_eth.c b/drivers/net/ethernet/xscale/ixp4xx_eth.c
index 758f820068b5..1e1779b53f22 100644
--- a/drivers/net/ethernet/xscale/ixp4xx_eth.c
+++ b/drivers/net/ethernet/xscale/ixp4xx_eth.c
@@ -250,6 +250,7 @@ static inline void memcpy_swab32(u32 *dest, u32 *src, int cnt)
 static DEFINE_SPINLOCK(mdio_lock);
 static struct eth_regs __iomem *mdio_regs; /* mdio command and status only */
 static struct mii_bus *mdio_bus;
+static struct device_node *mdio_bus_np;
 static int ports_open;
 static struct port *npe_port_tab[MAX_NPES];
 static struct dma_pool *dma_pool;
@@ -533,7 +534,8 @@ static int ixp4xx_mdio_register(struct eth_regs __iomem *regs)
 	mdio_bus->write = &ixp4xx_mdio_write;
 	snprintf(mdio_bus->id, MII_BUS_ID_SIZE, "ixp4xx-eth-0");
 
-	if ((err = mdiobus_register(mdio_bus)))
+	err = of_mdiobus_register(mdio_bus, mdio_bus_np);
+	if (err)
 		mdiobus_free(mdio_bus);
 	return err;
 }
@@ -1364,7 +1366,6 @@ static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
 	struct device_node *np = dev->of_node;
 	struct of_phandle_args queue_spec;
 	struct eth_plat_info *plat;
-	struct device_node *phy_np;
 	struct device_node *mdio_np;
 	u32 val;
 	int ret;
@@ -1381,25 +1382,12 @@ static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
 	/* NPE ID 0x00, 0x10, 0x20... */
 	plat->npe = (val << 4);
 
-	phy_np = of_parse_phandle(np, "phy-handle", 0);
-	if (phy_np) {
-		ret = of_property_read_u32(phy_np, "reg", &val);
-		if (ret) {
-			dev_err(dev, "cannot find phy reg\n");
-			return NULL;
-		}
-		of_node_put(phy_np);
-	} else {
-		dev_err(dev, "cannot find phy instance\n");
-		val = 0;
-	}
-	plat->phy = val;
-
 	/* Check if this device has an MDIO bus */
 	mdio_np = of_get_child_by_name(np, "mdio");
 	if (mdio_np) {
 		plat->has_mdio = true;
-		of_node_put(mdio_np);
+		mdio_bus_np = mdio_np;
+		/* DO NOT put the mdio_np, it will be used */
 	}
 
 	/* Get the rx queue as a resource from queue manager */
@@ -1539,10 +1527,14 @@ static int ixp4xx_eth_probe(struct platform_device *pdev)
 	__raw_writel(DEFAULT_CORE_CNTRL, &port->regs->core_control);
 	udelay(50);
 
-	snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
-		 mdio_bus->id, plat->phy);
-	phydev = phy_connect(ndev, phy_id, &ixp4xx_adjust_link,
-			     PHY_INTERFACE_MODE_MII);
+	if (np) {
+		phydev = of_phy_get_and_connect(ndev, np, ixp4xx_adjust_link);
+	} else {
+		snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
+			 mdio_bus->id, plat->phy);
+		phydev = phy_connect(ndev, phy_id, &ixp4xx_adjust_link,
+				     PHY_INTERFACE_MODE_MII);
+	}
 	if (!phydev) {
 		err = -ENODEV;
 		dev_err(dev, "no phydev\n");
-- 
2.29.2

Re: [PATCH 3/3] net: ethernet: ixp4xx: Use OF MDIO bus registration

[Linus Walleij]

On Tue, Apr 20, 2021 at 12:51 AM Linus Walleij <linus.walleij@linaro.org> wrote:

> This augments the IXP4xx to use the OF MDIO bus code when
> registering the MDIO bus and when looking up the PHY
> for the ethernet network device.
>
> Cc: Zoltan HERPAI <wigyori@uid0.hu>
> Cc: Raylynn Knight <rayknight@me.com>
> Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
> ---
> ChangeLog v1->v2:
> - New patch making use of more OF infrastructure.

Realized after sending that the MDIO people wanna look at this
too so adding them to CC, tell me if you want me to resend the
patch.

Yours,
Linus Walleij

Re: [PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[Andrew Lunn]

> +      mdio {
> +        #address-cells = <1>;
> +        #size-cells = <0>;
> +        phy1: phy@1 {
> +          #phy-cells = <0>;

Hi Linus

phy-cells is not part of the Ethernet PHY binding.

	  Andrew

Re: [PATCH 2/3] net: ethernet: ixp4xx: Support device tree probing

[Andrew Lunn]

> +	phy_np = of_parse_phandle(np, "phy-handle", 0);
> +	if (phy_np) {
> +		ret = of_property_read_u32(phy_np, "reg", &val);
> +		if (ret) {
> +			dev_err(dev, "cannot find phy reg\n");
> +			return NULL;
> +		}
> +		of_node_put(phy_np);
> +	} else {
> +		dev_err(dev, "cannot find phy instance\n");
> +		val = 0;
> +	}
> +	plat->phy = val;

phy-handle can point to a PHY on any bus. You should not assume it is
the devices own bus. Once you have phy_np call of_phy_find_device()
which gives you the actual phy device. Please don't let the
limitations of the current platform data limit you from implementing
this correctly.

      Andrew

Re: [PATCH 3/3] net: ethernet: ixp4xx: Use OF MDIO bus registration

[Andrew Lunn]

> @@ -1381,25 +1382,12 @@ static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
>  	/* NPE ID 0x00, 0x10, 0x20... */
>  	plat->npe = (val << 4);
>  
> -	phy_np = of_parse_phandle(np, "phy-handle", 0);
> -	if (phy_np) {
> -		ret = of_property_read_u32(phy_np, "reg", &val);
> -		if (ret) {
> -			dev_err(dev, "cannot find phy reg\n");
> -			return NULL;
> -		}
> -		of_node_put(phy_np);
> -	} else {
> -		dev_err(dev, "cannot find phy instance\n");
> -		val = 0;
> -	}
> -	plat->phy = val;
> -

Isn't this code you just added in the previous patch?

>  	/* Check if this device has an MDIO bus */
>  	mdio_np = of_get_child_by_name(np, "mdio");
>  	if (mdio_np) {
>  		plat->has_mdio = true;
> -		of_node_put(mdio_np);
> +		mdio_bus_np = mdio_np;
> +		/* DO NOT put the mdio_np, it will be used */
>  	}
>  
>  	/* Get the rx queue as a resource from queue manager */
> @@ -1539,10 +1527,14 @@ static int ixp4xx_eth_probe(struct platform_device *pdev)
>  	__raw_writel(DEFAULT_CORE_CNTRL, &port->regs->core_control);
>  	udelay(50);
>  
> -	snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
> -		 mdio_bus->id, plat->phy);
> -	phydev = phy_connect(ndev, phy_id, &ixp4xx_adjust_link,
> -			     PHY_INTERFACE_MODE_MII);
> +	if (np) {
> +		phydev = of_phy_get_and_connect(ndev, np, ixp4xx_adjust_link);
> +	} else {
> +		snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
> +			 mdio_bus->id, plat->phy);

mdiobus_get_phy() and phy_connect_direct() might be better.

	  Andrew

Re: [PATCH 2/3] net: ethernet: ixp4xx: Support device tree probing

[Linus Walleij]

On Tue, Apr 20, 2021 at 3:35 AM Andrew Lunn <andrew@lunn.ch> wrote:

> > +     phy_np = of_parse_phandle(np, "phy-handle", 0);
> > +     if (phy_np) {
> > +             ret = of_property_read_u32(phy_np, "reg", &val);
> > +             if (ret) {
> > +                     dev_err(dev, "cannot find phy reg\n");
> > +                     return NULL;
> > +             }
> > +             of_node_put(phy_np);
> > +     } else {
> > +             dev_err(dev, "cannot find phy instance\n");
> > +             val = 0;
> > +     }
> > +     plat->phy = val;
>
> phy-handle can point to a PHY on any bus. You should not assume it is
> the devices own bus. Once you have phy_np call of_phy_find_device()
> which gives you the actual phy device. Please don't let the
> limitations of the current platform data limit you from implementing
> this correctly.

In patch 3/3 I migrate to of_phy_get_and_connect() which I assume
fixes this, I just wanted to split up the work, but do you prefer
that I simply squash patches 2 & 3 into one?

Yours,
Linus Walleij

Re: [PATCH 3/3] net: ethernet: ixp4xx: Use OF MDIO bus registration

[Linus Walleij]

On Tue, Apr 20, 2021 at 3:47 AM Andrew Lunn <andrew@lunn.ch> wrote:

> > @@ -1381,25 +1382,12 @@ static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
> >       /* NPE ID 0x00, 0x10, 0x20... */
> >       plat->npe = (val << 4);
> >
> > -     phy_np = of_parse_phandle(np, "phy-handle", 0);
> > -     if (phy_np) {
> > -             ret = of_property_read_u32(phy_np, "reg", &val);
> > -             if (ret) {
> > -                     dev_err(dev, "cannot find phy reg\n");
> > -                     return NULL;
> > -             }
> > -             of_node_put(phy_np);
> > -     } else {
> > -             dev_err(dev, "cannot find phy instance\n");
> > -             val = 0;
> > -     }
> > -     plat->phy = val;
> > -
>
> Isn't this code you just added in the previous patch?

Yep. It's by the token of "one technical step per patch" but I
suggested that maybe you prefer to take two technical
steps in one big patch, then we can just squash
patches 2 & 3. I'll fix it as you want it just tell me how :)

> > -     snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
> > -              mdio_bus->id, plat->phy);
> > -     phydev = phy_connect(ndev, phy_id, &ixp4xx_adjust_link,
> > -                          PHY_INTERFACE_MODE_MII);
> > +     if (np) {
> > +             phydev = of_phy_get_and_connect(ndev, np, ixp4xx_adjust_link);
> > +     } else {
> > +             snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
> > +                      mdio_bus->id, plat->phy);
>
> mdiobus_get_phy() and phy_connect_direct() might be better.

Do you mean for the legacy code path (else clause), or the
new code path with of_phy_get_and_connect() or both?

I tried not to change the legacy code in order to not introduce
regressions, so if I change that I suppose it should be a
separate patch.

On the other hand this driver has not been much maintained
the recent years so we might need to be a bit rough when
bringing it into shape. (After migrating all of IXP4xx to
device tree a lot of the legacy code will eventually be deleted.)

Yours,
Linus Walleij

Re: [PATCH 3/3] net: ethernet: ixp4xx: Use OF MDIO bus registration

[Andrew Lunn]

On Tue, Apr 20, 2021 at 10:44:16AM +0200, Linus Walleij wrote:
> On Tue, Apr 20, 2021 at 3:47 AM Andrew Lunn <andrew@lunn.ch> wrote:
> 
> > > @@ -1381,25 +1382,12 @@ static struct eth_plat_info *ixp4xx_of_get_platdata(struct device *dev)
> > >       /* NPE ID 0x00, 0x10, 0x20... */
> > >       plat->npe = (val << 4);
> > >
> > > -     phy_np = of_parse_phandle(np, "phy-handle", 0);
> > > -     if (phy_np) {
> > > -             ret = of_property_read_u32(phy_np, "reg", &val);
> > > -             if (ret) {
> > > -                     dev_err(dev, "cannot find phy reg\n");
> > > -                     return NULL;
> > > -             }
> > > -             of_node_put(phy_np);
> > > -     } else {
> > > -             dev_err(dev, "cannot find phy instance\n");
> > > -             val = 0;
> > > -     }
> > > -     plat->phy = val;
> > > -
> >
> > Isn't this code you just added in the previous patch?
> 
> Yep. It's by the token of "one technical step per patch"

I don't actually seeing it being a step, since it is actually broken
and of_phy_get_and_connect() does pretty much everything it should do,
which is what you replace it with.

It is a long time since i converted a platform_data driver to DT. But
i remember trying to fill in the platform_data structure from DT was
often the wrong way to do it. They contain different data, and you
cannot easily map one to the other. So you need to make bigger changes
to the probe function. You have two intermingled code paths, one
dealing with platform_data, and the other using DT.

I've not looked in detail, but i guess my first step would be, cleanly
register the MDIO bus. Second step would be to register the PHY. And
it might need some refactoring patches just to make it easier to
understand.

> > > -     snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
> > > -              mdio_bus->id, plat->phy);
> > > -     phydev = phy_connect(ndev, phy_id, &ixp4xx_adjust_link,
> > > -                          PHY_INTERFACE_MODE_MII);
> > > +     if (np) {
> > > +             phydev = of_phy_get_and_connect(ndev, np, ixp4xx_adjust_link);
> > > +     } else {
> > > +             snprintf(phy_id, MII_BUS_ID_SIZE + 3, PHY_ID_FMT,
> > > +                      mdio_bus->id, plat->phy);
> >
> > mdiobus_get_phy() and phy_connect_direct() might be better.
> 
> Do you mean for the legacy code path (else clause), or the
> new code path with of_phy_get_and_connect() or both?
> 
> I tried not to change the legacy code in order to not introduce
> regressions, so if I change that I suppose it should be a
> separate patch.

Yes, the legacy code. You don't often see this string parsing
method. And since you have the bus, and the index, you can directly go
to the PHY avoiding it all. A separate patch would be better.

   Andrew

Re: [PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[Linus Walleij]

On Tue, Apr 20, 2021 at 3:26 AM Andrew Lunn <andrew@lunn.ch> wrote:

> > +      mdio {
> > +        #address-cells = <1>;
> > +        #size-cells = <0>;
> > +        phy1: phy@1 {
> > +          #phy-cells = <0>;
>
> Hi Linus
>
> phy-cells is not part of the Ethernet PHY binding.

Nevertheless:

  CHECK   Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.example.dt.yaml
/var/linus/linux-nomadik/build-ixp4/Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.example.dt.yaml:
phy@1: '#phy-cells' is a required property
    From schema:
/home/linus/.local/lib/python3.9/site-packages/dtschema/schemas/phy/phy-provider.yaml

It has been hardcoded as required into the dtschema python package.
Looks like this:

properties:
  $nodename:
    pattern: "^(|usb-|usb2-|usb3-|pci-|pcie-|sata-)phy(@[0-9a-f,]+)*$"

  "#phy-cells": true

  phy-supply: true

required:
  - "#phy-cells"

additionalProperties: true

If this is wrong I bet Rob needs to hear about it.

Yours,
Linus Walleij

Re: [PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[Andrew Lunn]

On Thu, Apr 22, 2021 at 05:39:07PM +0200, Linus Walleij wrote:
> On Tue, Apr 20, 2021 at 3:26 AM Andrew Lunn <andrew@lunn.ch> wrote:
> 
> > > +      mdio {
> > > +        #address-cells = <1>;
> > > +        #size-cells = <0>;
> > > +        phy1: phy@1 {
> > > +          #phy-cells = <0>;
> >
> > Hi Linus
> >
> > phy-cells is not part of the Ethernet PHY binding.
> 
> Nevertheless:
> 
>   CHECK   Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.example.dt.yaml
> /var/linus/linux-nomadik/build-ixp4/Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.example.dt.yaml:
> phy@1: '#phy-cells' is a required property
>     From schema:
> /home/linus/.local/lib/python3.9/site-packages/dtschema/schemas/phy/phy-provider.yaml
> 
> It has been hardcoded as required into the dtschema python package.
> Looks like this:
> 
> properties:
>   $nodename:
>     pattern: "^(|usb-|usb2-|usb3-|pci-|pcie-|sata-)phy(@[0-9a-f,]+)*$"
> 
>   "#phy-cells": true
> 
>   phy-supply: true
> 
> required:
>   - "#phy-cells"
> 
> additionalProperties: true
> 
> If this is wrong I bet Rob needs to hear about it.

That is the wrong sort of PHY. That is a generic PHY, not a PHY, aka
Ethernet PHY. Maybe you need to change the label to ethernet-phy ?

	 Andrew

Re: [PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[Linus Walleij]

On Thu, Apr 22, 2021 at 5:56 PM Andrew Lunn <andrew@lunn.ch> wrote:
> On Thu, Apr 22, 2021 at 05:39:07PM +0200, Linus Walleij wrote:
> > On Tue, Apr 20, 2021 at 3:26 AM Andrew Lunn <andrew@lunn.ch> wrote:
> >
> > > > +      mdio {
> > > > +        #address-cells = <1>;
> > > > +        #size-cells = <0>;
> > > > +        phy1: phy@1 {
> > > > +          #phy-cells = <0>;
> > >
> > > Hi Linus
> > >
> > > phy-cells is not part of the Ethernet PHY binding.
> >
> > Nevertheless:
> >
> >   CHECK   Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.example.dt.yaml
> > /var/linus/linux-nomadik/build-ixp4/Documentation/devicetree/bindings/net/intel,ixp4xx-ethernet.example.dt.yaml:
> > phy@1: '#phy-cells' is a required property
> >     From schema:
> > /home/linus/.local/lib/python3.9/site-packages/dtschema/schemas/phy/phy-provider.yaml
> >
> > It has been hardcoded as required into the dtschema python package.
> > Looks like this:
> >
> > properties:
> >   $nodename:
> >     pattern: "^(|usb-|usb2-|usb3-|pci-|pcie-|sata-)phy(@[0-9a-f,]+)*$"
> >
> >   "#phy-cells": true
> >
> >   phy-supply: true
> >
> > required:
> >   - "#phy-cells"
> >
> > additionalProperties: true
> >
> > If this is wrong I bet Rob needs to hear about it.
>
> That is the wrong sort of PHY. That is a generic PHY, not a PHY, aka
> Ethernet PHY.

It is a bit confusing :D not to mention that the term
"phy" or "physical interface" as I suppose it is meant to
be understood is a bit ambiguous to begin with.

> Maybe you need to change the label to ethernet-phy ?

Yeah that works, I'll do that.

Yours,
Linus Walleij

Re: [PATCH 1/3] net: ethernet: ixp4xx: Add DT bindings

[Andrew Lunn]

> It is a bit confusing :D not to mention that the term
> "phy" or "physical interface" as I suppose it is meant to
> be understood is a bit ambiguous to begin with.

Yes. For a long time, phy meant Ethernet PHY. And then the generic PHY
framework came along, and decided to reuse the phy prefix, rather than
genphy, or something unique. And now we have confusion.

	Andrew