Skip to contentSkip to conversation

net: macb: set interface name based on DT aliases

PATCH In review Boris BREZILLON 3 participants 3 messages Mar 12, 2014 – May 6, 2014

Subsystems (3)
Network drivers Device tree bindings Atmel MACB Ethernet driver
lore
Share & tools
lore

Boris BREZILLON

author
#0 lore raw

PATCH 1/1 net: macb: set interface name based on DT aliases

Use aliases to set the interface name (ethX) instead of automatic assignement.

Signed-off-by Boris BREZILLON <[email protected]>

---

Hello Nicolas, Jean-Christophe,

This is an example on how we could set the ethernet interface id based on DT aliases.
As you can see this patch is not properly separated.

I'm still not happy with this approach for several reasons:
1) If another ethernet iface has already been registered with the same id the net dev registration will fail.
2) We bypass the ethernet macros/functions and directly use the net device functions which IMHO is not future proof.

Best Regards,

Boris

 arch/arm/boot/dts/sama5d36.dtsi     |    2 +-
 arch/arm/boot/dts/sama5d3_emac.dtsi |    4 ++++
 arch/arm/boot/dts/sama5d3_gmac.dtsi |    4 ++++
 drivers/net/ethernet/cadence/macb.c |   15 +++++++++++++--
 4 files changed, 22 insertions(+), 3 deletions(-)
arch/arm/boot/dts/sama5d36.dtsi +1 -1
index 6c31c26..db58cad 100644
@@ -8,8 +8,8 @@
  */
 #include "sama5d3.dtsi"
 #include "sama5d3_can.dtsi"
-#include "sama5d3_emac.dtsi"
 #include "sama5d3_gmac.dtsi"
+#include "sama5d3_emac.dtsi"
 #include "sama5d3_lcd.dtsi"
 #include "sama5d3_mci2.dtsi"
 #include "sama5d3_tcb1.dtsi"
arch/arm/boot/dts/sama5d3_emac.dtsi +4 -0
index fe2af92..f2c5da0 100644
@@ -11,6 +11,10 @@
 #include <dt-bindings/interrupt-controller/irq.h>
 
 / {
+	aliases {
+		eth1 = &macb1;
+	};
+
 	ahb {
 		apb {
 			pinctrl at fffff200 {
arch/arm/boot/dts/sama5d3_gmac.dtsi +4 -0
index a6cb050..05164de 100644
@@ -11,6 +11,10 @@
 #include <dt-bindings/interrupt-controller/irq.h>
 
 / {
+	aliases {
+		eth0 = &macb0;
+	};
+
 	ahb {
 		apb {
 			pinctrl at fffff200 {
drivers/net/ethernet/cadence/macb.c +13 -2
index 3190d38..b6f1f7c 100644
@@ -1795,7 +1795,7 @@ static int __init macb_probe(struct platform_device *pdev)
 {
 	struct macb_platform_data *pdata;
 	struct resource *regs;
-	struct net_device *dev;
+	struct net_device *dev = NULL;
 	struct macb *bp;
 	struct phy_device *phydev;
 	u32 config;

@@ -1819,7 +1819,18 @@ static int __init macb_probe(struct platform_device *pdev)
 	}
 
 	err = -ENOMEM;
-	dev = alloc_etherdev(sizeof(*bp));
+	if (pdev->dev.of_node) {
+		int id = of_alias_get_id(pdev->dev.of_node, "eth");
+		if (id >= 0) {
+			char name[IFNAMSIZ];
+			snprintf(name, sizeof(name), "eth%d", id);
+			dev = alloc_netdev(sizeof(*bp), name, ether_setup);
+		}
+	}
+
+	if (!dev)
+		dev = alloc_etherdev(sizeof(*bp));
+
 	if (!dev)
 		goto err_out;
Signature · 1 line
1.7.9.5
#1 lore raw
Quoted 12 lines · Boris BREZILLON

Use aliases to set the interface name (ethX) instead of automatic assignement.

Signed-off-by Boris BREZILLON <[email protected]>

---

Hello Nicolas, Jean-Christophe,

This is an example on how we could set the ethernet interface id based on DT aliases.
As you can see this patch is not properly separated.

Quoted 5 lines

I'm still not happy with this approach for several reasons:
1) If another ethernet iface has already been registered with the same id the net dev registration will fail.
2) We bypass the ethernet macros/functions and directly use the net device functions which IMHO is not future proof.

[Sorry for the late reply, going through old backlog of unread email]

Looks like this should live in networking core instead of in the driver, it's how we handle it in other subsystems.

I don't know what DaveM's opinions on that is though.

-Olof

#2 lore raw

Hi Olof,

Quoted 18 lines · Olof Johansson
Quoted 12 lines · Boris BREZILLON

Use aliases to set the interface name (ethX) instead of automatic assignement.

Signed-off-by Boris BREZILLON <[email protected]>

---

Hello Nicolas, Jean-Christophe,

This is an example on how we could set the ethernet interface id based on DT aliases.
As you can see this patch is not properly separated.

Quoted 5 lines

I'm still not happy with this approach for several reasons:
1) If another ethernet iface has already been registered with the same id the net dev registration will fail.
2) We bypass the ethernet macros/functions and directly use the net device functions which IMHO is not future proof.

Quoted 4 lines

[Sorry for the late reply, going through old backlog of unread email]

Looks like this should live in networking core instead of in the driver, it's how we handle it in other subsystems.

Sure.
This version was just a quick an dirty hack, but I'll send a new version integrating this functionality in netdev core.

Quoted 1 line

I don't know what DaveM's opinions on that is though.

I'll put him in Cc of the next version.

Best Regards,

Boris