mx6cuboxi: Fix Ethernet PHY detection problem

PATCH New Fabio Estevam 3 participants 7 messages May 4, 2015

lore

Share & tools

lore

Fabio Estevam

author

2015-05-04 03:30 UTC #0 lore raw

PATCH 1/2 mx6cuboxi: Fix Ethernet PHY detection problem

From: Fabio Estevam <[email protected]>

mx6cuboxi sometimes fails to recognize the Ethernet PHY:

Net: Phy 0 not found

The explanation comes from a patch from Rabeeh:

"The LED_ACT pin on the carrier-one boards had a pull down that forces the phy address to 0x0; where on CuBox-i and the production HummingBoard that pin is connected directly to LED that depending on the pull down strength of the LED it might be sampled as '0' or '1' thus the phy address might appear as either address 0x0 or 0x4."

So fix this problem by telling phy_find_by_mask() to scan the PHY at addresses 0x0 and 0x4.

Reported-by Vagrant Cascadian <[email protected]>

Signed-off-by Rabeeh Khoury <[email protected]>

Signed-off-by Fabio Estevam <[email protected]>

---
 board/solidrun/mx6cuboxi/mx6cuboxi.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)

board/solidrun/mx6cuboxi/mx6cuboxi.c +30 -1

index 9aa0259..ff018e1 100644
@@ -25,6 +25,7 @@
 #include <asm/imx-common/video.h>
 #include <mmc.h>
 #include <fsl_esdhc.h>
+#include <malloc.h>
 #include <miiphy.h>
 #include <netdev.h>
 #include <asm/arch/crm_regs.h>

@@ -152,9 +153,14 @@ int board_phy_config(struct phy_device *phydev)
 	return 0;
 }
 
+/* On Cuboxi Ethernet PHY can be located at addresses 0x0 or 0x4 */
+#define ETH_PHY_MASK	((1 << 0x0) | (1 << 0x4))
+
 int board_eth_init(bd_t *bis)
 {
 	struct iomuxc *const iomuxc_regs = (struct iomuxc *)IOMUXC_BASE_ADDR;
+	struct mii_dev *bus;
+	struct phy_device *phydev;
 
 	int ret = enable_fec_anatop_clock(ENET_25MHZ);
 	if (ret)

@@ -165,7 +171,30 @@ int board_eth_init(bd_t *bis)
 
 	setup_iomux_enet();
 
-	return cpu_eth_init(bis);
+	bus = fec_get_miibus(IMX_FEC_BASE, -1);
+	if (!bus)
+		return -EINVAL;
+
+	phydev = phy_find_by_mask(bus, ETH_PHY_MASK, PHY_INTERFACE_MODE_RGMII);
+	if (!phydev) {
+		ret = -EINVAL;
+		goto free_bus;
+	}
+
+	debug("using phy@address %d\n", phydev->addr);
+	ret = fec_probe(bis, -1, IMX_FEC_BASE, bus, phydev);
+	if (ret) {
+		printf("FEC MXC: %s:failed\n", __func__);
+		goto free_phydev;
+	}
+
+	return 0;
+
+free_phydev:
+	free(phydev);
+free_bus:
+	free(bus);
+	return ret;
 }
 
 #ifdef CONFIG_VIDEO_IPUV3

Signature · 1 line

1.9.1

Fabio Estevam

author

2015-05-04 03:30 UTC #1 lore raw

PATCH 2/2 mx6cuboxi: Pull down PAD_ENET_RXD0/RXD1

From: Fabio Estevam <[email protected]>

Configure PAD_ENET_RXD0/RXD1 pads as pull down because these pads are directly connected to the Atheros 8035/8030 although they should be functional only in the RMII mode - 8030.

Signed-off-by Rabeeh Khoury <[email protected]>

Signed-off-by Fabio Estevam <[email protected]>

---
 board/solidrun/mx6cuboxi/mx6cuboxi.c | 2 ++
 1 file changed, 2 insertions(+)

board/solidrun/mx6cuboxi/mx6cuboxi.c +2 -0

index ff018e1..8fb6004 100644
@@ -134,6 +134,8 @@ static iomux_v3_cfg_t const enet_pads[] = {
 	IOMUX_PADS(PAD_RGMII_RD2__RGMII_RD2 | MUX_PAD_CTRL(ENET_PAD_CTRL)),
 	IOMUX_PADS(PAD_RGMII_RD3__RGMII_RD3 | MUX_PAD_CTRL(ENET_PAD_CTRL)),
 	IOMUX_PADS(PAD_RGMII_RX_CTL__RGMII_RX_CTL | MUX_PAD_CTRL(ENET_PAD_CTRL_PD)),
+	IOMUX_PADS(PAD_ENET_RXD0__GPIO1_IO27 | MUX_PAD_CTRL(ENET_PAD_CTRL_PD)),
+	IOMUX_PADS(PAD_ENET_RXD1__GPIO1_IO26 | MUX_PAD_CTRL(ENET_PAD_CTRL_PD)),
 };
 
 static void setup_iomux_enet(void)

Signature · 1 line

1.9.1

Nikolay Dimitrov

2015-05-04 04:18 UTC #2 lore raw

PATCH 1/2 mx6cuboxi: Fix Ethernet PHY detection problem

In reply to Fabio Estevam

Hi Fabio,

Quoted 6 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>
mx6cuboxi sometimes fails to recognize the Ethernet PHY:
Net: Phy 0 not found

Quoted 7 lines

The explanation comes from a patch from Rabeeh:
"The LED_ACT pin on the carrier-one boards had a pull down that forces the phy address to 0x0; where on CuBox-i and the production HummingBoard that pin is connected directly to LED that depending on the pull down strength of the LED it might be sampled as '0' or '1' thus the phy address might appear as either address 0x0 or 0x4."

There's no such thing as "LED pull-down". The forward voltage drop of a LED is between 1.65V (red low-power LEDs) to 2.1V (green LEDs) to even more for blue LEDs. Even the lowest Vf doesn't qualify as logic "0" for LVCMOS33, which is around 1V max (Vil). The LED just can't pull-down the voltage level low enough.

So, unless you have some control over the pin (via a programmable on-chip pull-up or pull-down) which I doubt as it's a PHY pin, the actual behavior is that the pin is floating, and samples a random value at boot. Which means, the hardware is just buggy.

Kind regards,
Nikolay

Fabio Estevam

author

2015-05-04 12:22 UTC #3 lore raw

PATCH 1/2 mx6cuboxi: Fix Ethernet PHY detection problem

Hi Nikolay,

Quoted 25 lines · Nikolay Dimitrov

Hi Fabio,
Quoted 7 lines · Fabio Estevam
From: Fabio Estevam <[email protected]>
mx6cuboxi sometimes fails to recognize the Ethernet PHY:
Net: Phy 0 not found
Quoted 7 lines
The explanation comes from a patch from Rabeeh:
"The LED_ACT pin on the carrier-one boards had a pull down that forces the phy address to 0x0; where on CuBox-i and the production HummingBoard that pin is connected directly to LED that depending on the pull down strength of the LED it might be sampled as '0' or '1' thus the phy address might appear as either address 0x0 or 0x4."
There's no such thing as "LED pull-down". The forward voltage drop of a LED is between 1.65V (red low-power LEDs) to 2.1V (green LEDs) to even more for blue LEDs. Even the lowest Vf doesn't qualify as logic "0" for LVCMOS33, which is around 1V max (Vil). The LED just can't pull-down the voltage level low enough.

Quoted 4 lines

So, unless you have some control over the pin (via a programmable on-chip pull-up or pull-down) which I doubt as it's a PHY pin, the actual behavior is that the pin is floating, and samples a random value at boot. Which means, the hardware is just buggy.

As mentioned in the commit log this explanation comes from Solid-run.

The key point here is that the PHY can appear at 0x0 and 0x4, so this patch handles such case.

Regards,

Fabio Estevam

Nikolay Dimitrov

2015-05-04 14:06 UTC #4 lore raw

PATCH 1/2 mx6cuboxi: Fix Ethernet PHY detection problem

Hi Fabio,

Quoted 35 lines · Fabio Estevam

Hi Nikolay,
Quoted 26 lines · Nikolay Dimitrov
Hi Fabio,
Quoted 7 lines · Fabio Estevam
From: Fabio Estevam <[email protected]>
mx6cuboxi sometimes fails to recognize the Ethernet PHY:
Net: Phy 0 not found
Quoted 8 lines
The explanation comes from a patch from Rabeeh:
"The LED_ACT pin on the carrier-one boards had a pull down that forces the phy address to 0x0; where on CuBox-i and the production HummingBoard that pin is connected directly to LED that depending on the pull down strength of the LED it might be sampled as '0' or '1' thus the phy address might appear as either address 0x0 or 0x4."
There's no such thing as "LED pull-down". The forward voltage drop of a LED is between 1.65V (red low-power LEDs) to 2.1V (green LEDs) to even more for blue LEDs. Even the lowest Vf doesn't qualify as logic "0" for LVCMOS33, which is around 1V max (Vil). The LED just can't pull-down the voltage level low enough.
Quoted 4 lines
So, unless you have some control over the pin (via a programmable on-chip pull-up or pull-down) which I doubt as it's a PHY pin, the
actual behavior is that the pin is floating, and samples a random
value at boot. Which means, the hardware is just buggy.

Quoted 5 lines

As mentioned in the commit log this explanation comes from Solid-run.
The key point here is that the PHY can appear at 0x0 and 0x4, so this patch handles such case.

Yes, I saw that. Sorry for the off-topic. The reason I allowed myself to comment is that this text will go into git log, and people can treat it as the proper way to configure boot-strapable pins, which I don't think it is.

Otherwise your patch is completely OK - this is the only way to fix such behavior of the hardware.

Kind regards,
Nikolay

Rabeeh Khoury

2015-05-04 12:35 UTC #5 lore raw

PATCH 1/2 mx6cuboxi: Fix Ethernet PHY detection problem

In reply to Nikolay Dimitrov

Quoted 17 lines · Nikolay Dimitrov

Hi Fabio,
Quoted 6 lines · Fabio Estevam
From: Fabio Estevam <[email protected]>
mx6cuboxi sometimes fails to recognize the Ethernet PHY:
Net: Phy 0 not found
Quoted 7 lines
The explanation comes from a patch from Rabeeh:
"The LED_ACT pin on the carrier-one boards had a pull down that forces the phy address to 0x0; where on CuBox-i and the production HummingBoard that pin is connected directly to LED that depending on the pull down strength of the LED it might be sampled as '0' or '1' thus the phy address might appear as either address 0x0 or 0x4."

Quoted 5 lines

There's no such thing as "LED pull-down". The forward voltage drop of a LED is between 1.65V (red low-power LEDs) to 2.1V (green LEDs) to even more for blue LEDs. Even the lowest Vf doesn't qualify as logic "0" for LVCMOS33, which is around 1V max (Vil). The LED just can't pull-down the voltage level low enough.

This is AR8035 implementation; in reset stage the LED pin is configured as input, and pull up/down does matter. In this case it configures the PHY address.

After reset is deasserted the same LED pin becomes output and then according to the previous pull/up it should be active high/low (i.e. driver or sink).

Nikolay Dimitrov

2015-05-04 13:49 UTC #6 lore raw

PATCH 1/2 mx6cuboxi: Fix Ethernet PHY detection problem

Hi Rabeeh,

Quoted 24 lines · Rabeeh Khoury

Quoted 18 lines · Nikolay Dimitrov
Hi Fabio,
Quoted 6 lines · Fabio Estevam
From: Fabio Estevam <[email protected]>
mx6cuboxi sometimes fails to recognize the Ethernet PHY:
Net: Phy 0 not found
Quoted 8 lines
The explanation comes from a patch from Rabeeh:
"The LED_ACT pin on the carrier-one boards had a pull down that forces the phy address to 0x0; where on CuBox-i and the production HummingBoard that pin is connected directly to LED that depending on the pull down strength of the LED it might be sampled as '0' or '1' thus the phy address might appear as either address 0x0 or 0x4."
Quoted 5 lines
There's no such thing as "LED pull-down". The forward voltage drop of a LED is between 1.65V (red low-power LEDs) to 2.1V (green LEDs) to even more for blue LEDs. Even the lowest Vf doesn't qualify as logic "0" for LVCMOS33, which is around 1V max (Vil). The LED just can't pull-down the voltage level low enough.

Quoted 7 lines

This is AR8035 implementation; in reset stage the LED pin is configured as input, and pull up/down does matter. In this case it configures the PHY address.
After reset is deasserted the same LED pin becomes output and then according to the previous pull/up it should be active high/low (i.e.
driver or sink).

I know this. This is why I said that the pin is floating and the R+LED can't define a proper logic level during configuration bootstrapping. In such situations there should be an external pull-up/pull-down to make sure things work properly.

Regards,
Nikolay