arm64: add ethernet to orange pi 3

Hello

2 sunxi board still does not have ethernet working, orangepi 1+ and orangepi 3.
This is due to the fact thoses boards have a PHY which need 2 regulators.

A first attempt by Ondřej Jirman was made to support them was made by adding support in stmmac driver:
https://lore.kernel.org/lkml/[email protected]/
Proposal rejected, since regulators need to be handled by the PHY core.

My first tentative was to just add handling of phy and phy-io in phy-core:
https://lore.kernel.org/netdev/[email protected]/T/
But having hard-coded phy names was rejected.

Second tentative tryed the same path than clocks and clock-names for regulators.
https://lore.kernel.org/netdev/[email protected]/t/
But using this was rejected by DT maintainers.

So v3 use a new regulator_bulk_get_all() which grab all supplies properties in a DT node.
But this way could have some problem, a netdev driver could handle already its PHY (like dwmac-sun8i already do) and so both phy-core and the netdev will use both.
It is why phy-supply was renamed in ephy-supply in patch #3.

This serie was tested on whole range of board and PHY architecture:
- internal PHY
* sun8i-h3-orangepi-pc
- external PHY
* sun50i-h6-pine-h64
* sun8i-r40-bananapi-m2-ultra
* sun8i-a83t-bananapi-m3
* sun50i-a64-bananapi-m64
* sun50i-h6-orangepi-3
* sun50i-h5-nanopi-neo-plus2

The resume/suspend of PHY was tested.

Regards

changes since v1:
- Add regulator_bulk_get_all for ease handling of PHY regulators
- Removed all convertion patchs to keep DT compatibility.

Changes since v2:
- removed use of regulator-names and regulators list.

Changes since v3:
- fixes kbuild robot report

Corentin Labbe (2):
regulator: Add of_regulator_bulk_get_all
phy: handle optional regulator for PHY

Ondřej Jirman (1):
arm64: dts: allwinner: orange-pi-3: Enable ethernet

 .../dts/allwinner/sun50i-h6-orangepi-3.dts    | 38 ++++++++
 drivers/net/mdio/fwnode_mdio.c                | 31 ++++++-
 drivers/net/phy/phy_device.c                  | 10 ++
 drivers/regulator/of_regulator.c              | 92 +++++++++++++++++++
 include/linux/phy.h                           |  3 +
 include/linux/regulator/consumer.h            |  8 ++
 6 files changed, 181 insertions(+), 1 deletion(-)

2.37.4

It work exactly like regulator_bulk_get() but instead of working on a provided list of names, it seek all consumers properties matching xxx-supply.

---
 drivers/regulator/of_regulator.c   | 92 ++++++++++++++++++++++++++++++
 include/linux/regulator/consumer.h |  8 +++
 2 files changed, 100 insertions(+)

index 0aff1c2886b5..584c92f1a317 100644
@@ -701,3 +701,95 @@ struct regulator_dev *of_parse_coupled_regulator(struct regulator_dev *rdev,
 
 	return c_rdev;
 }
+
+/*
+ * Check if name is a supply name according to the '*-supply' pattern
+ * return 0 if false
+ * return length of supply name without the -supply
+ */
+static int is_supply_name(const char *name)
+{
+	int strs, i;
+
+	strs = strlen(name);
+	/* string need to be at minimum len(x-supply) */
+	if (strs < 8)
+		return 0;
+	for (i = strs - 6; i > 0; i--) {
+		/* find first '-' and check if right part is supply */
+		if (name[i] != '-')
+			continue;
+		if (strcmp(name + i + 1, "supply") != 0)
+			return 0;
+		return i;
+	}
+	return 0;
+}
+
+/*
+ * of_regulator_bulk_get_all - get multiple regulator consumers
+ *
+ * @dev:	Device to supply
+ * @np:		device node to search for consumers
+ * @consumers:  Configuration of consumers; clients are stored here.
+ *
+ * @return number of regulators on success, an errno on failure.
+ *
+ * This helper function allows drivers to get several regulator
+ * consumers in one operation.  If any of the regulators cannot be
+ * acquired then any regulators that were allocated will be freed
+ * before returning to the caller.
+ */
+int of_regulator_bulk_get_all(struct device *dev, struct device_node *np,
+			      struct regulator_bulk_data **consumers)
+{
+	int num_consumers = 0;
+	struct regulator *tmp;
+	struct property *prop;
+	int i, n = 0, ret;
+	char name[64];
+
+	*consumers = NULL;
+
+	/*
+	 * first pass: get numbers of xxx-supply
+	 * second pass: fill consumers
+	 */
+restart:
+	for_each_property_of_node(np, prop) {
+		i = is_supply_name(prop->name);
+		if (i == 0)
+			continue;
+		if (!*consumers) {
+			num_consumers++;
+			continue;
+		} else {
+			memcpy(name, prop->name, i);
+			name[i] = '\0';
+			tmp = regulator_get(dev, name);
+			if (!tmp) {
+				ret = -EINVAL;
+				goto error;
+			}
+			(*consumers)[n].consumer = tmp;
+			n++;
+			continue;
+		}
+	}
+	if (*consumers)
+		return num_consumers;
+	if (num_consumers == 0)
+		return 0;
+	*consumers = kmalloc_array(num_consumers,
+				   sizeof(struct regulator_bulk_data),
+				   GFP_KERNEL);
+	if (!*consumers)
+		return -ENOMEM;
+	goto restart;
+
+error:
+	while (--n >= 0)
+		regulator_put(consumers[n]->consumer);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(of_regulator_bulk_get_all);

index 628a52b8e63f..39b666b40ea6 100644
@@ -244,6 +244,8 @@ int regulator_disable_deferred(struct regulator *regulator, int ms);
 
 int __must_check regulator_bulk_get(struct device *dev, int num_consumers,
 				    struct regulator_bulk_data *consumers);
+int __must_check of_regulator_bulk_get_all(struct device *dev, struct device_node *np,
+					   struct regulator_bulk_data **consumers);
 int __must_check devm_regulator_bulk_get(struct device *dev, int num_consumers,
 					 struct regulator_bulk_data *consumers);
 void devm_regulator_bulk_put(struct regulator_bulk_data *consumers);

@@ -481,6 +483,12 @@ static inline int devm_regulator_bulk_get(struct device *dev, int num_consumers,
 	return 0;
 }
 
+static inline int of_regulator_bulk_get_all(struct device *dev, struct device_node *np,
+					    struct regulator_bulk_data **consumers)
+{
+	return 0;
+}
+
 static inline int regulator_bulk_enable(int num_consumers,
 					struct regulator_bulk_data *consumers)
 {

2.37.4

What's the use case - why would a device not know which supplies it requires? This just looks like an invitation to badly written consumers TBH.

Hello

The device know which supply it have, but I found only this way to made all maintainers happy.
See https://lore.kernel.org/netdev/[email protected]/t/#m7a2e012f4c7c7058478811929774ab2af9bfcbf6

Regards

Well, it's not making this maintainer happy :/ If we know what PHY is there why not just look up the set of supplies based on the compatible of the PHY?

It looks to me like this series fetches the regulators before the PHY is bound to the driver, so what you're proposing would mean that the core PHY code would need a table of all compatibles (which is pretty hard to do, they encode the vendor/device ID, not some descriptive name) and then a list of the regulator names. IMHO that doesn't scale.

RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!

Oh, PHYs have interesting enough drivers to dynamically load here? The last time I was looking at MDIO stuff it was all running from generic class devices but that was quite a while ago.

There's a couple of generic drivers which are used if there isn't a specific driver available for the vendor/device ID that has either been probed from the hardware, or discovered encoded in the firmware's compatible property.

RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!

Yes. And you sometimes have the chicken/egg problem that you don't know what PHY it is until you have turned its regulators on and you can talk to it. So the PHY code will poke around in the DT description, and turn on the regulator before enumerating the bus.

Andrew

The following changes since commit 9abf2313adc1ca1b6180c508c25f22f9395cc780:

Linux 6.1-rc1 (2022-10-16 15:36:24 -0700)

are available in the Git repository at:

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator.git tags/of_regulator_bulk_get_all

for you to fetch changes up to 27b9ecc7a9ba1d0014779bfe5a6dbf630899c6e7:

regulator: Add of_regulator_bulk_get_all (2022-11-18 15:13:34 +0000)

----------------------------------------------------------------
regulator: Add of_regulator_bulk_get_all()

This adds a new of_regulator_bulk_get_all() which grab all supplies properties in a DT node, for use in implementing generic handling for things like MDIO PHYs where the physical standardisation of the bus does not extend to power supplies.

----------------------------------------------------------------
Corentin Labbe (1):
regulator: Add of_regulator_bulk_get_all

 drivers/regulator/of_regulator.c   | 92 ++++++++++++++++++++++++++++++++++++++
 include/linux/regulator/consumer.h |  8 ++++
 2 files changed, 100 insertions(+)

Add handling of optional regulators for PHY. Regulators need to be enabled before PHY scanning, so MDIO bus initiate this task.

---
 drivers/net/mdio/fwnode_mdio.c | 31 ++++++++++++++++++++++++++++++-
 drivers/net/phy/phy_device.c   | 10 ++++++++++
 include/linux/phy.h            |  3 +++
 3 files changed, 43 insertions(+), 1 deletion(-)

index 689e728345ce..19a16072d4ca 100644
@@ -10,6 +10,7 @@
 #include <linux/fwnode_mdio.h>
 #include <linux/of.h>
 #include <linux/phy.h>
+#include <linux/regulator/consumer.h>
 #include <linux/pse-pd/pse.h>
 
 MODULE_AUTHOR("Calvin Johnson <[email protected]>");

@@ -116,7 +117,9 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
 	struct phy_device *phy;
 	bool is_c45 = false;
 	u32 phy_id;
-	int rc;
+	int rc, reg_cnt = 0;
+	struct regulator_bulk_data *consumers = NULL;
+	struct device_node __maybe_unused *nchild = NULL;
 
 	psec = fwnode_find_pse_control(child);
 	if (IS_ERR(psec))

@@ -133,6 +136,26 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
 	if (rc >= 0)
 		is_c45 = true;
 
+#ifdef CONFIG_OF
+	for_each_child_of_node(bus->dev.of_node, nchild) {
+		u32 reg;
+
+		of_property_read_u32(nchild, "reg", &reg);
+		if (reg != addr)
+			continue;
+		reg_cnt = of_regulator_bulk_get_all(&bus->dev, nchild, &consumers);
+		if (reg_cnt > 0) {
+			rc = regulator_bulk_enable(reg_cnt, consumers);
+			if (rc)
+				return rc;
+		}
+		if (reg_cnt < 0) {
+			dev_err(&bus->dev, "Fail to regulator_bulk_get_all err=%d\n", reg_cnt);
+			return reg_cnt;
+		}
+	}
+#endif
+
 	if (is_c45 || fwnode_get_phy_id(child, &phy_id))
 		phy = get_phy_device(bus, addr, is_c45);
 	else

@@ -142,6 +165,9 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
 		goto clean_mii_ts;
 	}
 
+	phy->regulator_cnt = reg_cnt;
+	phy->consumers = consumers;
+
 	if (is_acpi_node(child)) {
 		phy->irq = bus->irq[addr];
 

@@ -180,6 +206,9 @@ int fwnode_mdiobus_register_phy(struct mii_bus *bus,
 clean_pse:
 	pse_control_put(psec);
 
+	if (reg_cnt > 0)
+		regulator_bulk_disable(reg_cnt, consumers);
+
 	return rc;
 }
 EXPORT_SYMBOL(fwnode_mdiobus_register_phy);

index 57849ac0384e..957e27c75eb2 100644
@@ -28,6 +28,7 @@
 #include <linux/phy_led_triggers.h>
 #include <linux/pse-pd/pse.h>
 #include <linux/property.h>
+#include <linux/regulator/consumer.h>
 #include <linux/sfp.h>
 #include <linux/skbuff.h>
 #include <linux/slab.h>

@@ -1818,6 +1819,9 @@ int phy_suspend(struct phy_device *phydev)
 	if (!ret)
 		phydev->suspended = true;
 
+	if (phydev->regulator_cnt > 0)
+		regulator_bulk_disable(phydev->regulator_cnt, phydev->consumers);
+
 	return ret;
 }
 EXPORT_SYMBOL(phy_suspend);

@@ -1844,6 +1848,12 @@ int phy_resume(struct phy_device *phydev)
 {
 	int ret;
 
+	if (phydev->regulator_cnt > 0) {
+		ret = regulator_bulk_enable(phydev->regulator_cnt, phydev->consumers);
+		if (ret)
+			return ret;
+	}
+
 	mutex_lock(&phydev->lock);
 	ret = __phy_resume(phydev);
 	mutex_unlock(&phydev->lock);

index 9a3752c0c444..5d1311b35cc3 100644
@@ -729,6 +729,9 @@ struct phy_device {
 	void (*phy_link_change)(struct phy_device *phydev, bool up);
 	void (*adjust_link)(struct net_device *dev);
 
+	int regulator_cnt;
+	struct regulator_bulk_data *consumers;
+
 #if IS_ENABLED(CONFIG_MACSEC)
 	/* MACsec management functions */
 	const struct macsec_ops *macsec_ops;

2.37.4

From: Ondřej Jirman <[email protected]>

Orange Pi 3 has two regulators that power the Realtek RTL8211E PHY. According to the datasheet, both regulators need to be enabled at the same time, or that "phy-io" should be enabled slightly earlier than "phy" regulator.

RTL8211E/RTL8211EG datasheet says:

Note 4: 2.5V (or 1.8/1.5V) RGMII power should be risen simultaneously or slightly earlier than 3.3V power. Rising 2.5V (or 1.8/1.5V) power later than 3.3V power may lead to errors.

The timing is set in DT via startup-delay-us.

---
 .../dts/allwinner/sun50i-h6-orangepi-3.dts    | 38 +++++++++++++++++++
 1 file changed, 38 insertions(+)

index 6fc65e8db220..081b004f26ea 100644
@@ -13,6 +13,7 @@ / {
 	compatible = "xunlong,orangepi-3", "allwinner,sun50i-h6";
 
 	aliases {
+		ethernet0 = &emac;
 		serial0 = &uart0;
 		serial1 = &uart1;
 	};

@@ -55,6 +56,15 @@ led-1 {
 		};
 	};
 
+	reg_gmac_2v5: gmac-2v5 {
+		compatible = "regulator-fixed";
+		regulator-name = "gmac-2v5";
+		regulator-min-microvolt = <2500000>;
+		regulator-max-microvolt = <2500000>;
+		enable-active-high;
+		gpio = <&pio 3 6 GPIO_ACTIVE_HIGH>; /* PD6 */
+	};
+
 	reg_vcc5v: vcc5v {
 		/* board wide 5V supply directly from the DC jack */
 		compatible = "regulator-fixed";

@@ -113,6 +123,33 @@ &ehci3 {
 	status = "okay";
 };
 
+&emac {
+	pinctrl-names = "default";
+	pinctrl-0 = <&ext_rgmii_pins>;
+	phy-mode = "rgmii-id";
+	phy-handle = <&ext_rgmii_phy>;
+	status = "okay";
+};
+
+&mdio {
+	ext_rgmii_phy: ethernet-phy@1 {
+		compatible = "ethernet-phy-ieee802.3-c22";
+		reg = <1>;
+		/*
+		 * The board uses 2.5V RGMII signalling. Power sequence to enable
+		 * the phy is to enable GMAC-2V5 and GMAC-3V (aldo2) power rails
+		 * at the same time and to wait 100ms. The driver enables phy-io
+		 * first. Delay is achieved with enable-ramp-delay on reg_aldo2.
+		 */
+		phy-io-supply = <&reg_gmac_2v5>;
+		ephy-supply = <&reg_aldo2>;
+
+		reset-gpios = <&pio 3 14 GPIO_ACTIVE_LOW>; /* PD14 */
+		reset-assert-us = <15000>;
+		reset-deassert-us = <40000>;
+	};
+};
+
 &gpu {
 	mali-supply = <&reg_dcdcc>;
 	status = "okay";

@@ -211,6 +248,7 @@ reg_aldo2: aldo2 {
 				regulator-min-microvolt = <3300000>;
 				regulator-max-microvolt = <3300000>;
 				regulator-name = "vcc33-audio-tv-ephy-mac";
+				regulator-enable-ramp-delay = <100000>;
 			};
 
 			/* ALDO3 is shorted to CLDO1 */

2.37.4

A MAC driver will put its DT properties in the MAC node. A PHY will put its DT properties in the PHY node of the MDIO bus. Since they are in different locations, they can use the same name. So please keep with phy-supply.

Please also update Documentation/devicetree/bindings/net/ethernet-phy.yaml with these new properties.

Andrew

Applied to

broonie/regulator.git for-next

Thanks!

[1/3] regulator: Add of_regulator_bulk_get_all

      commit: 27b9ecc7a9ba1d0014779bfe5a6dbf630899c6e7

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark