Skip to contentSkip to conversation

drivers/of/of_mdio.c:fix of_mdiobus_register()

PATCH In review Dajun Jin 4 participants 8 messages Mar 1, 2020 – Mar 4, 2020

Subsystems (3)
lore
Share & tools
lore

Dajun Jin

author
#0 lore raw

PATCH 1/1 drivers/of/of_mdio.c:fix of_mdiobus_register()

when registers a phy_device successful, should terminate the loop or the phy_device would be registered in other addr.

---
 drivers/of/of_mdio.c | 1 +
 1 file changed, 1 insertion(+)
drivers/of/of_mdio.c +1 -0
index 8270bbf505fb..9f982c0627a0 100644
@@ -306,6 +306,7 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np)
 				rc = of_mdiobus_register_phy(mdio, child, addr);
 				if (rc && rc != -ENODEV)
 					goto unregister;
+				break;
 			}
 		}
 	}
Signature · 1 line
2.17.1
#1 lore raw
Quoted 13 lines · Dajun Jin

when registers a phy_device successful, should terminate the loop or the phy_device would be registered in other addr.

---
 drivers/of/of_mdio.c | 1 +
 1 file changed, 1 insertion(+)
drivers/of/of_mdio.c +0 -0
index 8270bbf505fb..9f982c0627a0 100644

drivers/of/of_mdio.c int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np) · line 312

Quoted 7 lines
drivers/of/of_mdio.c +1 -0
@@ -306,6 +306,7 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np)
 				rc = of_mdiobus_register_phy(mdio, child, addr);
 				if (rc && rc != -ENODEV)
 					goto unregister;
+				break;
 			}
 		}
 	}

Hi Dajun

What problem are you seeing? You explanation needs to be better.

I'm guessing you have two or more PHYs on the bus, without reg properties?

Andrew

author
#2 lore raw
Quoted 24 lines · Dajun Jin
Quoted 13 lines

when registers a phy_device successful, should terminate the loop or the phy_device would be registered in other addr.

---
 drivers/of/of_mdio.c | 1 +
 1 file changed, 1 insertion(+)
drivers/of/of_mdio.c +0 -0
index 8270bbf505fb..9f982c0627a0 100644

drivers/of/of_mdio.c int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np) · line 312

Quoted 7 lines
drivers/of/of_mdio.c +1 -0
@@ -306,6 +306,7 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np)
 				rc = of_mdiobus_register_phy(mdio, child, addr);
 				if (rc && rc != -ENODEV)
 					goto unregister;
+				break;
 			}
 		}
 	}

Hi Dajun

Quoted 6 lines

What problem are you seeing? You explanation needs to be better.

I'm guessing you have two or more PHYs on the bus, without reg properties?

Andrew

Hi Andrew

If a phy without reg property would be registered to all unoccupied addr.

This is my test in Xilinx zcu106 board.

dts is liks this:

ethernet@ff0e0000 {
    compatible = "cdns,zynqmp-gem", "cdns,gem";
    status = "okay";
    ...
    
    phy@0 {
        ti,rx-internal-delay = <0x8>;
        ti,tx-internal-delay = <0xa>;
        ti,fifo-depth = <0x1>;
        ti,rxctrl-strap-worka;
        linux,phandle = <0x12>;
        phandle = <0x12>;
    };
};

then when borad is booting,the dmesg is like this:

[    4.600035] mdio_bus ff0e0000.ethernet-ffffffff: /amba/ethernet@ff0e0000/phy@0 has invalid PHY address
[    4.600050] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 0
[    4.602076] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 1
[    4.603849] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 2
[    4.605574] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 4
[    4.607312] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 5
...
[    4.636155] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 28
[    4.637335] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 29
[    4.638504] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 30
[    4.639666] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 31

Dajun

#3 lore raw

Hi Dajun

Quoted 19 lines

This is my test in Xilinx zcu106 board.

dts is liks this:

ethernet@ff0e0000 {
    compatible = "cdns,zynqmp-gem", "cdns,gem";
    status = "okay";
    ...
    
    phy@0 {
        ti,rx-internal-delay = <0x8>;
        ti,tx-internal-delay = <0xa>;
        ti,fifo-depth = <0x1>;
        ti,rxctrl-strap-worka;
        linux,phandle = <0x12>;
        phandle = <0x12>;
    };
};

then when borad is booting,the dmesg is like this:

Quoted 11 lines
[    4.600035] mdio_bus ff0e0000.ethernet-ffffffff: /amba/ethernet@ff0e0000/phy@0 has invalid PHY address
[    4.600050] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 0
[    4.602076] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 1
[    4.603849] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 2
[    4.605574] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 4
[    4.607312] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 5
...
[    4.636155] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 28
[    4.637335] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 29
[    4.638504] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 30
[    4.639666] mdio_bus ff0e0000.ethernet-ffffffff: scan phy phy at address 31

For a single PHY without a reg properties, it should do the right thing. But it will go wrong if there are multiple PHYs without reg properties. In that case, the break helps.

However, as the comment suggests, you really should have a reg property.

Please make the commit message better, and then i will give a reviewed-by.

Thanks
Andrew

#4 lore raw
Quoted 17 lines · Andrew Lunn

Hi Dajun

Quoted 15 lines

This is my test in Xilinx zcu106 board.

dts is liks this:

ethernet@ff0e0000 {
    compatible = "cdns,zynqmp-gem", "cdns,gem";
    status = "okay";
    ...
    
    phy@0 {
        ti,rx-internal-delay = <0x8>;
        ti,tx-internal-delay = <0xa>;
        ti,fifo-depth = <0x1>;
        ti,rxctrl-strap-worka;
        linux,phandle = <0x12>;
        phandle = <0x12>;

Isn't dtc going to complain about this? The node name has an address, but there's no reg property. If there's no reg property, shouldn't it be just "phy" ?

The above doesn't look like the original .dts file itself either, but a .dtb translated back to a .dts - note the numeric phandle properties and presence of "linux,phandle".

ethernet-phy.yaml also says:

required:
- reg

so arguably the above doesn't conform to what we expect?

Signature · 3 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up
According to speedtest.net: 11.9Mbps down 500kbps up
author
#5 lore raw

PATCH 1/1 drivers/of/of_mdio.c:fix of_mdiobus_register()

In reply to Andrew Lunn

When registers a phy_device successful, should terminate the loop or the phy_device would be registered in other addr. If there are multiple PHYs without reg properties, it will go wrong.

---
Hi Andrew, Thanks for your review.

 drivers/of/of_mdio.c | 1 +
 1 file changed, 1 insertion(+)
drivers/of/of_mdio.c +1 -0
index 8270bbf505fb..9f982c0627a0 100644
@@ -306,6 +306,7 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np)
 				rc = of_mdiobus_register_phy(mdio, child, addr);
 				if (rc && rc != -ENODEV)
 					goto unregister;
+				break;
 			}
 		}
 	}
Signature · 1 line
2.17.1