Skip to contentSkip to conversation

Marvell MDIO driver and new MDIO name property

PATCH In review Alex Marginean 4 participants 12 messages Jul 25, 2019 – Sep 4, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v2 cover Marvell MDIO driver and new MDIO name property

This is a respin of the original Marvell MDIO driver submitted by Ken about one year ago, using the new MDIO DM code.
Added binding documents for mdio and marvell-mdio and introduced a new optional property for naming MDIO buses in case the driver can't reasonably pick a nice name by itself.

Changes in v2: add "-mdio" to device-name for cpm/cps

Alex Marginean (4):
net: mdio-uclass: name MDIO according to device-name property if
preset
doc: bindings: add mdio.txt describing generic MDIO properties
drivers: net: add marvell MDIO driver
arm: dts: Set custom names for cp110 master/slave MDIO buses

 arch/arm/dts/armada-cp110-master.dtsi         |   1 +
 arch/arm/dts/armada-cp110-slave.dtsi          |   1 +
 doc/device-tree-bindings/net/marvell-mdio.txt |  15 ++
 doc/device-tree-bindings/net/mdio.txt         |  36 +++
 drivers/net/Kconfig                           |  10 +
 drivers/net/Makefile                          |   1 +
 drivers/net/mvmdio.c                          | 236 ++++++++++++++++++
 net/mdio-uclass.c                             |  11 +
 8 files changed, 311 insertions(+)
 create mode 100644 doc/device-tree-bindings/net/marvell-mdio.txt
 create mode 100644 doc/device-tree-bindings/net/mdio.txt
 create mode 100644 drivers/net/mvmdio.c
Signature · 1 line
2.17.1
#1 lore raw

PATCH v2 1/4 net: mdio-uclass: name MDIO according to device-name property if preset

Use the optional property device-name to name the MDIO bus. This works around limitations with using the DT node name on devices such as Armada-8040, which integrates two cp100 cores, both featuring MDIOs at the same relative offsets and with the same DT node names.
The concept was originally proposed by Marvell as a custom property called mdio-name specific to Marvell driver. This patch uses the more generic property device-name and moves this into MDIO class code so other can use it as well.

---
 net/mdio-uclass.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
net/mdio-uclass.c +11 -0
index 36a404ff44..428f3a994f 100644
@@ -23,6 +23,17 @@ void dm_mdio_probe_devices(void)
 
 static int dm_mdio_post_bind(struct udevice *dev)
 {
+	const char *dt_name;
+
+	/* set a custom name for the MDIO device, if present in DT */
+	if (ofnode_valid(dev->node)) {
+		dt_name = ofnode_read_string(dev->node, "device-name");
+		if (dt_name) {
+			debug("renaming dev %s to %s\n", dev->name, dt_name);
+			device_set_name(dev, dt_name);
+		}
+	}
+
 	/*
 	 * MDIO command doesn't like spaces in names, don't allow them to keep
 	 * it happy
Signature · 1 line
2.17.1
#3 lore raw

PATCH v2 2/4 doc: bindings: add mdio.txt describing generic MDIO properties

In reply to Alex Marginean

Adds a binding document for mdio. A notable deviation from corresponding Linux binding is the introduction of device-name optional property, which can be used to name MDIO buses. Two reset optional properties described by Linux binding are also not present as they don't seem to be used in U-Boot at this time.

---
 doc/device-tree-bindings/net/mdio.txt | 36 +++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 doc/device-tree-bindings/net/mdio.txt
doc/device-tree-bindings/net/mdio.txt new file +36 -0
new file mode 100644
index 0000000000..1595325050
@@ -0,0 +1,36 @@
+Common MDIO bus properties.
+
+These are generic properties that can apply to any MDIO bus.
+
+Optional properties:
+	- device-name - If present it is used to name the device and MDIO bus.
+			The name must be unique and must not contain spaces.
+
+A list of child nodes, one per device on the bus is expected.  These could be
+PHYs, switches or similar devices and child nodes should follow the specific
+binding for the device type.
+
+Example :
+This example shows the structure used for the external MDIO bus on NXP LS1028A
+RDB board.  Note that this MDIO device is an integrated PCI function and
+requires no compatible property for probing.
+
+/* definition in SoC dtsi file */
+	pcie at 1f0000000 {
+
+		mdio0: pci at 0,3 {
+			#address-cells=<0>;
+			#size-cells=<1>;
+			reg = <0x000300 0 0 0 0>;
+			status = "disabled";
+			device-name = "emdio";
+		};
+	};
+/* definition of PHYs in RDB dts file */
+&mdio0 {
+	status = "okay";
+	rdb_phy0: phy at 2 {
+		reg = <2>;
+	};
+};
+
Signature · 1 line
2.17.1
#4 lore raw

PATCH v2 2/4 doc: bindings: add mdio.txt describing generic MDIO properties

I added the optional property name in order that the legacy miiphy_get_dev_by_name() could be used with the mdio name.

And I think that my mdio searching way was much better than linux at least at last year.
I do not know now how in linux net interface finds its phy's mdio bus now. But at last year, linux stored all phys' FDT node addesses under all mdio buses, and when searching network interface' phy, linux driver compared its phy fdt node addresses to each phy fdt node address it stored one by one. So I added mdio_device_get_from_phy/mdio_mii_bus_get_from_phy/mdio_device_get_from_eth to simply the mdio/phy searching.

Yours,
Ken

-----Original Message-----
From: Alex Marginean <[email protected]>
Sent: Thursday, July 25, 2019 11:33 AM
To: u-boot at lists.denx.de
Cc: Joe Hershberger <[email protected]>; Ken Ma <[email protected]>; Nevo Hed <[email protected]>; Alex Marginean <[email protected]>; Alex Marginean <[email protected]>
Subject: [EXT] [PATCH v2 2/4] doc: bindings: add mdio.txt describing generic MDIO properties

External Email

----------------------------------------------------------------------
Adds a binding document for mdio. A notable deviation from corresponding Linux binding is the introduction of device-name optional property, which can be used to name MDIO buses. Two reset optional properties described by Linux binding are also not present as they don't seem to be used in U-Boot at this time.

---
 doc/device-tree-bindings/net/mdio.txt | 36 +++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 doc/device-tree-bindings/net/mdio.txt
doc/device-tree-bindings/net/mdio.txt new file +36 -0
new file mode 100644
index 0000000000..1595325050
@@ -0,0 +1,36 @@
+Common MDIO bus properties.
+
+These are generic properties that can apply to any MDIO bus.
+
+Optional properties:
+	- device-name - If present it is used to name the device and MDIO bus.
+			The name must be unique and must not contain spaces.
+
+A list of child nodes, one per device on the bus is expected.  These 
+could be PHYs, switches or similar devices and child nodes should 
+follow the specific binding for the device type.
+
+Example :
+This example shows the structure used for the external MDIO bus on NXP 
+LS1028A RDB board.  Note that this MDIO device is an integrated PCI 
+function and requires no compatible property for probing.
+
+/* definition in SoC dtsi file */
+	pcie at 1f0000000 {
+
+		mdio0: pci at 0,3 {
+			#address-cells=<0>;
+			#size-cells=<1>;
+			reg = <0x000300 0 0 0 0>;
+			status = "disabled";
+			device-name = "emdio";
+		};
+	};
+/* definition of PHYs in RDB dts file */
+&mdio0 {
+	status = "okay";
+	rdb_phy0: phy at 2 {
+		reg = <2>;
+	};
+};
+

--
2.17.1

#5 lore raw

PATCH v2 2/4 doc: bindings: add mdio.txt describing generic MDIO properties

Quoted 2 lines · Ken Ma

I added the optional property name in order that the legacy miiphy_get_dev_by_name() could be used with the mdio name.

That works with these patches, miiphy_get_dev_by_name is used, among others, by mdio command and mdio command shows the device with the name from DT, if the name property exists in DT.
There are two differences. First is property name which is now device-name instead of mdio-name, I used that hoping that it caches on and other classes would use the same.
The other difference is that the device-name binding is now part of mdio, rather than marvell-mdio, and the code is in the uclass rather than MDIO driver so all can use it.

Quoted 6 lines

And I think that my mdio searching way was much better than linux at
least at last year. > I do not know now how in linux net interface finds its phy's mdio bus now. But at last year, linux stored all phys' FDT node addesses under
all mdio buses, and when searching network interface' phy, linux
driver compared its phy fdt node addresses to each phy fdt node
address it stored one by one.

OK, you're probably looking at of_phy_find_device. It does search PHYs by DT node, which is useful for instance when using a phy-handle property in the ethernet node. I suppose your comment is that the search function could be simpler and I could agree with that. You could alternatively use mdiobus_get_phy which takes in a PHY address and doesn't care about DT nodes, but this only works if the PHY was probed already, I think. But that's all Linux implementation, it doesn't mean U-Boot has to do the exact same thing.

Quoted 4 lines

So I added mdio_device_get_from_phy/mdio_mii_bus_get_from_phy/mdio_device_get_from_eth to simply the mdio/phy searching. >
Yours,
Ken

For instance you would have a phy-handle property, eth driver would call mdio_device_get_from_phy on the phy node to get the MDIO bus and then call dm_mdio_phy_connect with the PHY address, correct?

I am planning to add a helper that actually does all that in one go, not just return a reference to MDIO bus based on PHY node or eth interface.

Wwe could have something similar to of_phy_get_and_connect in Linux. It would take in the ethernet udevice, read phy-handle from associated DT node, get PHY address and MDIO bus from DT and create the new PHY device. Right now I added that code to our eth driver (see [1]), but given that bindings are generic it would be nice to move all that code out into generic code as a helper function. The API would be something like this:

struct phy_device *dm_eth_phy_connect(struct udevice *eth); returning a phy device or null if something is missing.

It would be nice to also support fixed link bindings in there, but that is off-topic.

Does that look reasonable to you?
Thanks!
Alex

[1] enetc_start_phy in https://patchwork.ozlabs.org/patch/1126769/

Quoted 66 lines

-----Original Message-----
From: Alex Marginean <[email protected]>
Sent: Thursday, July 25, 2019 11:33 AM
To: u-boot at lists.denx.de
Cc: Joe Hershberger <[email protected]>; Ken Ma <[email protected]>; Nevo Hed <[email protected]>; Alex Marginean <[email protected]>; Alex Marginean <[email protected]>
Subject: [EXT] [PATCH v2 2/4] doc: bindings: add mdio.txt describing generic MDIO properties

External Email

----------------------------------------------------------------------
Adds a binding document for mdio. A notable deviation from corresponding Linux binding is the introduction of device-name optional property, which can be used to name MDIO buses. Two reset optional properties described by Linux binding are also not present as they don't seem to be used in U-Boot at this time.

---
  doc/device-tree-bindings/net/mdio.txt | 36 +++++++++++++++++++++++++++
  1 file changed, 36 insertions(+)

create mode 100644 doc/device-tree-bindings/net/mdio.txt

doc/device-tree-bindings/net/mdio.txt new file +36 -1
new file mode 100644
index 0000000000..1595325050
@@ -0,0 +1,36 @@
+Common MDIO bus properties.
+
+These are generic properties that can apply to any MDIO bus.
+
+Optional properties:
+	- device-name - If present it is used to name the device and MDIO bus.
+			The name must be unique and must not contain spaces.
+
+A list of child nodes, one per device on the bus is expected.  These
+could be PHYs, switches or similar devices and child nodes should
+follow the specific binding for the device type.
+
+Example :
+This example shows the structure used for the external MDIO bus on NXP
+LS1028A RDB board.  Note that this MDIO device is an integrated PCI
+function and requires no compatible property for probing.
+
+/* definition in SoC dtsi file */
+	pcie at 1f0000000 {
+
+		mdio0: pci at 0,3 {
+			#address-cells=<0>;
+			#size-cells=<1>;
+			reg = <0x000300 0 0 0 0>;
+			status = "disabled";
+			device-name = "emdio";
+		};
+	};
+/* definition of PHYs in RDB dts file */
+&mdio0 {
+	status = "okay";
+	rdb_phy0: phy at 2 {
+		reg = <2>;
+	};
+};
+
--

2.17.1

#6 lore raw

PATCH v2 2/4 doc: bindings: add mdio.txt describing generic MDIO properties

Please see my inline rely, thanks a lot!

Quoted 2 lines · Ken Ma

I added the optional property name in order that the legacy miiphy_get_dev_by_name() could be used with the mdio name.

That works with these patches, miiphy_get_dev_by_name is used, among others, by mdio command and mdio command shows the device with the name from DT, if the name property exists in DT. There are two differences. First is property name which is now device-name instead of mdio-name, I used that hoping that it caches on and other classes would use the same. The other difference is that the device-name binding is now part of mdio, rather than marvell-mdio, and the code is in the uclass rather than MDIO driver so all can use it. ###[Ken] Thank you very much for your detailed explanation, I have not read uboot codes for almost 1 year, 😊

Quoted 6 lines

And I think that my mdio searching way was much better than linux at least at last year. > I do not know now how in linux net interface finds its phy's mdio bus now. But at last year, linux stored all phys' FDT node addesses under all mdio buses, and when searching network interface' phy, linux driver compared its phy fdt node addresses to each phy fdt node address it stored one by one.

OK, you're probably looking at of_phy_find_device. It does search PHYs by DT node, which is useful for instance when using a phy-handle property in the ethernet node. I suppose your comment is that the search function could be simpler and I could agree with that. You could alternatively use mdiobus_get_phy which takes in a PHY address and doesn't care about DT nodes, but this only works if the PHY was probed already, I think. But that's all Linux implementation, it doesn't mean U-Boot has to do the exact same thing. ###[Ken] I think Linux cycle searching has too low efficiency, 😊

Quoted 6 lines

So I added mdio_device_get_from_phy/mdio_mii_bus_get_from_phy/mdio_device_get_fro m_eth
to simply the mdio/phy searching. >
Yours,
Ken

For instance you would have a phy-handle property, eth driver would call mdio_device_get_from_phy on the phy node to get the MDIO bus and then call dm_mdio_phy_connect with the PHY address, correct?

I am planning to add a helper that actually does all that in one go, not just return a reference to MDIO bus based on PHY node or eth interface.

Wwe could have something similar to of_phy_get_and_connect in Linux. It would take in the ethernet udevice, read phy-handle from associated DT node, get PHY address and MDIO bus from DT and create the new PHY device. Right now I added that code to our eth driver (see [1]), but given that bindings are generic it would be nice to move all that code out into generic code as a helper function. The API would be something like this:

struct phy_device *dm_eth_phy_connect(struct udevice *eth); returning a phy device or null if something is missing.

It would be nice to also support fixed link bindings in there, but that is off-topic.
###[Ken] I had though this solution in my old design, but later I denied it because mdio_device_get_from_phy() should be called in eth device probe step while dm_mdio_phy_connect() should be called in eth device start step, eth probe and start are 2 different steps, 😊

Does that look reasonable to you?
Thanks!
Alex

[1] enetc_start_phy in https://patchwork.ozlabs.org/patch/1126769/

Quoted 71 lines

-----Original Message-----
From: Alex Marginean <[email protected]>
Sent: Thursday, July 25, 2019 11:33 AM
To: u-boot at lists.denx.de
Cc: Joe Hershberger <[email protected]>; Ken Ma <[email protected]>; Nevo Hed <[email protected]>; Alex Marginean <[email protected]>; Alex Marginean <[email protected]>
Subject: [EXT] [PATCH v2 2/4] doc: bindings: add mdio.txt describing generic MDIO properties

External Email

----------------------------------------------------------------------
Adds a binding document for mdio. A notable deviation from corresponding Linux binding is the introduction of device-name optional property, which can be used to name MDIO buses. Two reset optional properties described by Linux binding are also not present as they don't seem to be used in U-Boot at this time.

---
  doc/device-tree-bindings/net/mdio.txt | 36 +++++++++++++++++++++++++++
  1 file changed, 36 insertions(+)

create mode 100644 doc/device-tree-bindings/net/mdio.txt

doc/device-tree-bindings/net/mdio.txt new file +36 -1
new file mode 100644
index 0000000000..1595325050
@@ -0,0 +1,36 @@
+Common MDIO bus properties.
+
+These are generic properties that can apply to any MDIO bus.
+
+Optional properties:
+	- device-name - If present it is used to name the device and MDIO bus.
+			The name must be unique and must not contain spaces.
+
+A list of child nodes, one per device on the bus is expected.  These 
+could be PHYs, switches or similar devices and child nodes should 
+follow the specific binding for the device type.
+
+Example :
+This example shows the structure used for the external MDIO bus on 
+NXP LS1028A RDB board.  Note that this MDIO device is an integrated 
+PCI function and requires no compatible property for probing.
+
+/* definition in SoC dtsi file */
+	pcie at 1f0000000 {
+
+		mdio0: pci at 0,3 {
+			#address-cells=<0>;
+			#size-cells=<1>;
+			reg = <0x000300 0 0 0 0>;
+			status = "disabled";
+			device-name = "emdio";
+		};
+	};
+/* definition of PHYs in RDB dts file */
+&mdio0 {
+	status = "okay";
+	rdb_phy0: phy at 2 {
+		reg = <2>;
+	};
+};
+
--

2.17.1

#8 lore raw

PATCH v2 3/4 drivers: net: add marvell MDIO driver

In reply to Alex Marginean

This patch adds a separate driver for the MDIO interface of the Marvell Ethernet controllers based on driver model. There are two reasons to have a separate driver rather than including it inside the MAC driver itself:
*) The MDIO interface is shared by all Ethernet ports, so a driver
must guarantee non-concurrent accesses to this MDIO interface. The most logical way is to have a separate driver that handles this single MDIO interface, used by all Ethernet ports.
*) The MDIO interface is the same between the existing mv643xx_eth
driver and the new mvneta/mvpp2 driver. Even though it is for now only used by the mvneta/mvpp2 driver, it will in the future be used by the mv643xx_eth driver as well.

This driver supports SMI IEEE for 802.3 Clause 22 and XSMI for IEEE 802.3 Clause 45.

This patch also adds device tree binding for marvell MDIO driver.

---
 doc/device-tree-bindings/net/marvell-mdio.txt |  15 ++
 drivers/net/Kconfig                           |  10 +
 drivers/net/Makefile                          |   1 +
 drivers/net/mvmdio.c                          | 236 ++++++++++++++++++
 4 files changed, 262 insertions(+)
 create mode 100644 doc/device-tree-bindings/net/marvell-mdio.txt
 create mode 100644 drivers/net/mvmdio.c
doc/device-tree-bindings/net/marvell-mdio.txt new file +15 -0
new file mode 100644
index 0000000000..e2038e2145
@@ -0,0 +1,15 @@
+* Marvell MDIO Ethernet Controller interface
+
+The Ethernet controllers of the Marvel Armada 3700 and Armada 7k/8k
+have an identical unit that provides an interface with the MDIO bus.
+This driver handles this MDIO interface.
+
+Mandatory properties:
+SoC specific:
+	- #address-cells: Must be <1>.
+	- #size-cells: Must be <0>.
+	- compatible: Should be "marvell,orion-mdio" (for SMI)
+				"marvell,xmdio"	     (for XSMI)
+	- reg: Base address and size SMI/XMSI bus.
+
+Please refer to "mdio.txt" for generic MDIO bus bindings.
drivers/net/Kconfig +10 -0
index 4d85fb1716..a982bc3ac5 100644
@@ -595,4 +595,14 @@ config FSL_ENETC
 	  This driver supports the NXP ENETC Ethernet controller found on some
 	  of the NXP SoCs.
 
+config MVMDIO
+	bool "Marvell MDIO interface support"
+	depends on DM_MDIO
+	help
+	  This driver supports the MDIO interface found in the network
+	  interface units of the Marvell EBU SoCs (Kirkwood, Orion5x,
+	  Dove, Armada 370, Armada XP, Armada 37xx and Armada7K/8K/8KP).
+
+	  This driver is used by the MVPP2 and MVNETA drivers.
+
 endif # NETDEVICES
drivers/net/Makefile +1 -0
index 97119cec7c..5921fa3e9c 100644
@@ -80,3 +80,4 @@ obj-$(CONFIG_HIGMACV300_ETH) += higmacv300.o
 obj-$(CONFIG_MDIO_SANDBOX) += mdio_sandbox.o
 obj-$(CONFIG_FSL_ENETC) += fsl_enetc.o fsl_enetc_mdio.o
 obj-$(CONFIG_MDIO_MUX_SANDBOX) += mdio_mux_sandbox.o
+obj-$(CONFIG_MVMDIO) += mvmdio.o
drivers/net/mvmdio.c new file +236 -0

236 lines not on this page · raw message

Signature · 1 line
2.17.1
#10 lore raw

PATCH v2 4/4 arm: dts: Set custom names for cp110 master/slave MDIO buses

In reply to Alex Marginean

Implicitly Marvell MDIO driver uses DT node names for devices, but in this case that is not unique. Set MDIO device names for master/slave to cpm/cps.

---
 arch/arm/dts/armada-cp110-master.dtsi | 1 +
 arch/arm/dts/armada-cp110-slave.dtsi  | 1 +
 2 files changed, 2 insertions(+)
arch/arm/dts/armada-cp110-master.dtsi +1 -0
index 551d00d774..e4c17e9f4b 100644
@@ -96,6 +96,7 @@
 				#size-cells = <0>;
 				compatible = "marvell,orion-mdio";
 				reg = <0x12a200 0x10>;
+				device-name = "cpm-mdio";
 			};
 
 			cpm_syscon0: system-controller at 440000 {
arch/arm/dts/armada-cp110-slave.dtsi +1 -0
index 2ea9004f1d..2fbd7b5514 100644
@@ -96,6 +96,7 @@
 				#size-cells = <0>;
 				compatible = "marvell,orion-mdio";
 				reg = <0x12a200 0x10>;
+				device-name = "cps-mdio";
 			};
 
 			cps_syscon0: system-controller at 440000 {
Signature · 1 line
2.17.1