Skip to contentSkip to conversation

net: bcmgenet: Fix attaching to PYH failed on RPi 4B

PATCH Superseded Jian-Hong Pan 7 participants 19 messages Jun 21, 2021 – Jun 24, 2021

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net: bcmgenet: Fix attaching to PYH failed on RPi 4B

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

---
 drivers/net/ethernet/broadcom/genet/bcmmii.c | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)
drivers/net/ethernet/broadcom/genet/bcmmii.c +17 -2
index 5335244e4577..64f244471fd3 100644
@@ -289,6 +289,7 @@ int bcmgenet_mii_probe(struct net_device *dev)
 	struct phy_device *phydev;
 	u32 phy_flags = 0;
 	int ret;
+	int i;
 
 	/* Communicate the integrated PHY revision */
 	if (priv->internal_phy)
@@ -301,8 +302,22 @@ int bcmgenet_mii_probe(struct net_device *dev)
 	priv->old_pause = -1;
 
 	if (dn) {
-		phydev = of_phy_connect(dev, priv->phy_dn, bcmgenet_mii_setup,
-					phy_flags, priv->phy_interface);
+		/* Try to connect the PHY on UniMAC DMIO bus up to 3 times.
+		 * Wait a while between each time for mdio-bcm-unimac module's
+		 * loading and probing.
+		 */
+		phydev = NULL;
+		for (i = 1; i < 4 && !phydev; i++) {
+			netdev_info(dev,
+				    "connect %s on UniMAC MDIO bus %d time",
+				    priv->phy_dn->full_name, i);
+			phydev = of_phy_connect(dev, priv->phy_dn,
+						bcmgenet_mii_setup,
+						phy_flags, priv->phy_interface);
+			if (!phydev && i < 3)
+				msleep(500);
+		}
+
 		if (!phydev) {
 			pr_err("could not attach to PHY\n");
 			return -ENODEV;
Signature · 1 line
2.32.0
maintainer
#1 lore raw
Quoted 13 lines · Jian-Hong Pan

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Andrew

#2 lore raw
Quoted 14 lines · Andrew Lunn
Quoted 13 lines · Jian-Hong Pan

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 7 lines
Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

(quoted diff) +0 -0

b/drivers/net/mdio/mdio-bcm-unimac.c
index bfc9be23c973..d1844ef3724a 100644

drivers/net/mdio/mdio-bcm-unimac.c +1 -0
@@ -351,6 +351,7 @@ static struct platform_driver unimac_mdio_driver = {
                .pm = &unimac_mdio_pm_ops,
        },
        .probe  = unimac_mdio_probe,
+       .probe_type = PROBE_FORCE_SYNCHRONOUS,
        .remove = unimac_mdio_remove,
 };
 module_platform_driver(unimac_mdio_driver);
Signature · 1 line
Florian
#3 lore raw
Quoted 35 lines · Florian Fainelli
Quoted 14 lines · Andrew Lunn
Quoted 13 lines · Jian-Hong Pan

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 7 lines
Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

drivers/net/mdio/mdio-bcm-unimac.c +0 -0
index bfc9be23c973..d1844ef3724a 100644

drivers/net/mdio/mdio-bcm-unimac.c static struct platform_driver unimac_mdio_driver = { · line 357

Quoted 7 lines
drivers/net/mdio/mdio-bcm-unimac.c +1 -0
@@ -351,6 +351,7 @@ static struct platform_driver unimac_mdio_driver = {
                .pm = &unimac_mdio_pm_ops,
        },
        .probe  = unimac_mdio_probe,
+       .probe_type = PROBE_FORCE_SYNCHRONOUS,
        .remove = unimac_mdio_remove,
 };
 module_platform_driver(unimac_mdio_driver);

This won't build try this instead:

(quoted diff) +0 -0

b/drivers/net/mdio/mdio-bcm-unimac.c
index bfc9be23c973..53fecb53cd65 100644

drivers/net/mdio/mdio-bcm-unimac.c +1 -0
@@ -349,6 +349,7 @@ static struct platform_driver unimac_mdio_driver = {
                .name = UNIMAC_MDIO_DRV_NAME,
                .of_match_table = unimac_mdio_ids,
                .pm = &unimac_mdio_pm_ops,
+               .probe_type = PROBE_FORCE_SYNCHRONOUS,
        },
        .probe  = unimac_mdio_probe,
        .remove = unimac_mdio_remove,
Signature · 1 line
Florian
#4 lore raw
In reply to Florian Fainelli
Quoted 24 lines · Florian Fainelli
Quoted 14 lines · Andrew Lunn
Quoted 13 lines · Jian-Hong Pan

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 7 lines
Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Quoted 5 lines

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

For reference we have mdio_bcm_unimac/genet both built as modules in Fedora and I've not seen this issue reported using vanilla upstream kernels if that's a useful reference point.

Regards,
Peter

#5 lore raw

Am 21.06.21 um 18:56 schrieb Peter Robinson:

Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Andrew Lunn
Quoted 20 lines · Jian-Hong Pan
Quoted 13 lines

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Quoted 8 lines
Quoted 5 lines

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

For reference we have mdio_bcm_unimac/genet both built as modules in Fedora and I've not seen this issue reported using vanilla upstream kernels if that's a useful reference point.

I was also unable to reproduce this issue, but it seems to be a known issue [1], [2].

Jian-Hong opened an issue in my Github repo [3], but before the issue was narrowed down, he decided to send this workaround.

Best regards

[1] -
https://github.com/raspberrypi/linux/issues/3108#issuecomment-723550749
[2] - https://archlinuxarm.org/forum/viewtopic.php?f=67&t=14806
[3] - https://github.com/lategoodbye/rpi-zero/issues/51

Quoted 8 lines

Regards,
Peter

_______________________________________________
linux-rpi-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-rpi-kernel

#6 lore raw
Quoted 32 lines · Stefan Wahren

Am 21.06.21 um 18:56 schrieb Peter Robinson:

Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Andrew Lunn
Quoted 20 lines · Jian-Hong Pan
Quoted 13 lines

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Quoted 8 lines
Quoted 5 lines

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

For reference we have mdio_bcm_unimac/genet both built as modules in Fedora and I've not seen this issue reported using vanilla upstream kernels if that's a useful reference point.

Quoted 5 lines

I was also unable to reproduce this issue, but it seems to be a known issue [1], [2].

Jian-Hong opened an issue in my Github repo [3], but before the issue was narrowed down, he decided to send this workaround.

The comment about changing the phy-mode property is not quite making sense to me, except if that means that in one case the Broadcom PHY driver is used and in the other case the Generic PHY driver is used.

What is not clear to me from the debugging that has been done so far is whether the mdio-bcm-unimac MDIO controller was not loaded at the time of_phy_connect() was trying to identify the PHY device.
--
Florian

#7 lore raw

Florian Fainelli <[email protected]> 於 2021年6月22日 週二 上午5:47寫道:

Quoted 40 lines · Stefan Wahren
Quoted 32 lines

Am 21.06.21 um 18:56 schrieb Peter Robinson:

Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Andrew Lunn
Quoted 20 lines · Jian-Hong Pan
Quoted 13 lines

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Quoted 8 lines
Quoted 5 lines

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

For reference we have mdio_bcm_unimac/genet both built as modules in Fedora and I've not seen this issue reported using vanilla upstream kernels if that's a useful reference point.

Quoted 5 lines

I was also unable to reproduce this issue, but it seems to be a known issue [1], [2].

Jian-Hong opened an issue in my Github repo [3], but before the issue was narrowed down, he decided to send this workaround.

Quoted 7 lines

The comment about changing the phy-mode property is not quite making sense to me, except if that means that in one case the Broadcom PHY driver is used and in the other case the Generic PHY driver is used.

What is not clear to me from the debugging that has been done so far is whether the mdio-bcm-unimac MDIO controller was not loaded at the time of_phy_connect() was trying to identify the PHY device.

MODULE_SOFTDEP("pre: mdio-bcm-unimac") mentioned in the comment [1] solves this issue.

Tracing the code by following the debug message in comment #2 [2], I learned the path bcmgenet_mii_probe()'s of_phy_connect() -> of_phy_find_device() -> of_mdio_find_device() -> bus_find_device_by_of_node(). And, bus_find_device_by_of_node() cannot find the device on the mdio bus.

So, I traced bcm2711-rpi-4-b's device tree to find out which one is the mdio device and why it has not been prepared ready on the mdio bus for genet.
Then, I found out it is mdio-bcm-unimac module as mentioned in comment #4 [3]. Also, noticed "unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus" comes after "bcmgenet fd580000.ethernet eth0: failed to connect to PHY" in the log.

With these findings, I try to re-modprobe genet module again. The ethernet on RPi 4B works correctly! Also, noticed mdio-bcm-unimac module is loaded before I re-modprobe genet module. Therefore, I try to make mdio-bcm-unimac built in kernel image, instead of a module. Then, genet always can find the mdio device on the bus and the ethernet works as well.

Consequently, the idea, loading mdio-bcm-unimac module earlier than genet module comes in my head! However, I don't know the key word "MODULE_SOFTDEP" until Florian's guide. That is why I have a loop to connect the PHY in the original patch. But, I understand MODULE_SOFTDEP is a better solution now!

I think this is like the module loading order situation mentioned in commit 11287b693d03 ("r8169: load Realtek PHY driver module before r8169") [4].

[1] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c6
[2] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c2
[3] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c4
[4] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=11287b693d03830010356339e4ceddf47dee34fa

Jian-Hong Pan

maintainer
#8 lore raw
Quoted 79 lines · Jian-Hong Pan

Florian Fainelli <[email protected]> 於 2021年6月22日 週二 上午5:47寫道:

Quoted 40 lines · Stefan Wahren
Quoted 32 lines

Am 21.06.21 um 18:56 schrieb Peter Robinson:

Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Andrew Lunn
Quoted 20 lines · Jian-Hong Pan
Quoted 13 lines

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Quoted 8 lines
Quoted 5 lines

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

For reference we have mdio_bcm_unimac/genet both built as modules in Fedora and I've not seen this issue reported using vanilla upstream kernels if that's a useful reference point.

Quoted 5 lines

I was also unable to reproduce this issue, but it seems to be a known issue [1], [2].

Jian-Hong opened an issue in my Github repo [3], but before the issue was narrowed down, he decided to send this workaround.

Quoted 7 lines

The comment about changing the phy-mode property is not quite making sense to me, except if that means that in one case the Broadcom PHY driver is used and in the other case the Generic PHY driver is used.

What is not clear to me from the debugging that has been done so far is whether the mdio-bcm-unimac MDIO controller was not loaded at the time of_phy_connect() was trying to identify the PHY device.

MODULE_SOFTDEP("pre: mdio-bcm-unimac") mentioned in the comment [1] solves this issue.

Tracing the code by following the debug message in comment #2 [2], I learned the path bcmgenet_mii_probe()'s of_phy_connect() -> of_phy_find_device() -> of_mdio_find_device() -> bus_find_device_by_of_node(). And, bus_find_device_by_of_node() cannot find the device on the mdio bus.

So, I traced bcm2711-rpi-4-b's device tree to find out which one is the mdio device and why it has not been prepared ready on the mdio bus for genet.
Then, I found out it is mdio-bcm-unimac module as mentioned in comment #4 [3]. Also, noticed "unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus" comes after "bcmgenet fd580000.ethernet eth0: failed to connect to PHY" in the log.

With these findings, I try to re-modprobe genet module again. The ethernet on RPi 4B works correctly! Also, noticed mdio-bcm-unimac module is loaded before I re-modprobe genet module. Therefore, I try to make mdio-bcm-unimac built in kernel image, instead of a module. Then, genet always can find the mdio device on the bus and the ethernet works as well.

Consequently, the idea, loading mdio-bcm-unimac module earlier than genet module comes in my head! However, I don't know the key word "MODULE_SOFTDEP" until Florian's guide. That is why I have a loop to connect the PHY in the original patch. But, I understand MODULE_SOFTDEP is a better solution now!

Quoted 4 lines

I think this is like the module loading order situation mentioned in commit 11287b693d03 ("r8169: load Realtek PHY driver module before r8169") [4].

The reason in r8169 is different. When people add r8169 module to initramfs but not the Realtek PHY driver module then loading r8169 will fail. The MODULE_SOFTDEP is a hint to tools building initramfs.

Quoted 7 lines

[1] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c6
[2] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c2
[3] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c4
[4] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=11287b693d03830010356339e4ceddf47dee34fa

Jian-Hong Pan

#9 lore raw

Heiner Kallweit <[email protected]> 於 2021年6月22日 週二 下午2:50寫道:

Quoted 75 lines · Jian-Hong Pan
Quoted 73 lines

Florian Fainelli <[email protected]> 於 2021年6月22日 週二 上午5:47寫道:

Quoted 40 lines · Stefan Wahren
Quoted 32 lines

Am 21.06.21 um 18:56 schrieb Peter Robinson:

Quoted 22 lines · Florian Fainelli
Quoted 21 lines · Andrew Lunn
Quoted 20 lines · Jian-Hong Pan
Quoted 13 lines

The Broadcom UniMAC MDIO bus comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY.

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 3 lines

This patch makes GENET try to connect the PHY up to 3 times. Also, waits a while between each time for mdio-bcm-unimac module's loading and probing.

Don't loop. Return -EPROBE_DEFER. The driver core will then probed the driver again later, by which time, the MDIO bus driver should of probed.

Quoted 8 lines
Quoted 5 lines

This is unlikely to work because GENET register the mdio-bcm-unimac platform device so we will likely run into a chicken and egg problem, though surprisingly I have not seen this on STB platforms where GENET is used, I will try building everything as a module like you do. Can you see if the following helps:

For reference we have mdio_bcm_unimac/genet both built as modules in Fedora and I've not seen this issue reported using vanilla upstream kernels if that's a useful reference point.

Quoted 5 lines

I was also unable to reproduce this issue, but it seems to be a known issue [1], [2].

Jian-Hong opened an issue in my Github repo [3], but before the issue was narrowed down, he decided to send this workaround.

Quoted 7 lines

The comment about changing the phy-mode property is not quite making sense to me, except if that means that in one case the Broadcom PHY driver is used and in the other case the Generic PHY driver is used.

What is not clear to me from the debugging that has been done so far is whether the mdio-bcm-unimac MDIO controller was not loaded at the time of_phy_connect() was trying to identify the PHY device.

MODULE_SOFTDEP("pre: mdio-bcm-unimac") mentioned in the comment [1] solves this issue.

Tracing the code by following the debug message in comment #2 [2], I learned the path bcmgenet_mii_probe()'s of_phy_connect() -> of_phy_find_device() -> of_mdio_find_device() -> bus_find_device_by_of_node(). And, bus_find_device_by_of_node() cannot find the device on the mdio bus.

So, I traced bcm2711-rpi-4-b's device tree to find out which one is the mdio device and why it has not been prepared ready on the mdio bus for genet.
Then, I found out it is mdio-bcm-unimac module as mentioned in comment #4 [3]. Also, noticed "unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus" comes after "bcmgenet fd580000.ethernet eth0: failed to connect to PHY" in the log.

With these findings, I try to re-modprobe genet module again. The ethernet on RPi 4B works correctly! Also, noticed mdio-bcm-unimac module is loaded before I re-modprobe genet module. Therefore, I try to make mdio-bcm-unimac built in kernel image, instead of a module. Then, genet always can find the mdio device on the bus and the ethernet works as well.

Quoted 5 lines
Quoted 5 lines

Consequently, the idea, loading mdio-bcm-unimac module earlier than genet module comes in my head! However, I don't know the key word "MODULE_SOFTDEP" until Florian's guide. That is why I have a loop to connect the PHY in the original patch. But, I understand MODULE_SOFTDEP is a better solution now!

Forgot to place some reference as note:

* MODULE_SOFTDEP is defined in include/linux/module.h [1]

* modprobe.d has an example: [2]
Assume "softdep c pre: a b post: d e" is provided in the configuration. Running "modprobe c" is now equivalent to "modprobe a b c d e" without the softdep.

[1] https://elixir.bootlin.com/linux/v5.13-rc7/source/include/linux/module.h#L170
[2] https://man7.org/linux/man-pages/man5/modprobe.d.5.html

Quoted 8 lines
Quoted 4 lines

I think this is like the module loading order situation mentioned in commit 11287b693d03 ("r8169: load Realtek PHY driver module before r8169") [4].

The reason in r8169 is different. When people add r8169 module to initramfs but not the Realtek PHY driver module then loading r8169 will fail. The MODULE_SOFTDEP is a hint to tools building initramfs.

Thanks for Heiner's quick clarification.
Maybe I missed some background of the commit ("r8169: load Realtek PHY driver module before r8169").

Jian-Hong Pan

Quoted 8 lines
Quoted 7 lines

[1] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c6
[2] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c2
[3] https://bugzilla.kernel.org/show_bug.cgi?id=213485#c4
[4] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=11287b693d03830010356339e4ceddf47dee34fa

Jian-Hong Pan

#10 lore raw

PATCH v2 1/1 net: bcmgenet: Fix attaching to PYH failed on RPi 4B

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

This patch adds the soft dependency to load mdio-bcm-unimac module before genet module to avoid the issue.

---
v2: Load mdio-bcm-unimac before genet module instead of trying to
connect the PHY in a loop.

 drivers/net/ethernet/broadcom/genet/bcmgenet.c | 1 +
 1 file changed, 1 insertion(+)
drivers/net/ethernet/broadcom/genet/bcmgenet.c +1 -0
index fcca023f22e5..41f7f078cd27 100644
@@ -4296,3 +4296,4 @@ MODULE_AUTHOR("Broadcom Corporation");
 MODULE_DESCRIPTION("Broadcom GENET Ethernet controller driver");
 MODULE_ALIAS("platform:bcmgenet");
 MODULE_LICENSE("GPL");
+MODULE_SOFTDEP("pre: mdio-bcm-unimac");
Signature · 1 line
2.32.0
#11 lore raw

The subject should be fixed s/PYH/PHY/ and also probably reworded to be:

net: bcmgenet: Add mdio-bcm-unimac soft dependency

Quoted 1 line · Jian-Hong Pan

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late.

It is not just coming too late, there is also no way for the module loader to figure out the dependency between GENET and its MDIO bus driver unless we provide this MODULE_SOFTDEP hint.

Quoted 12 lines

So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 5 lines

This patch adds the soft dependency to load mdio-bcm-unimac module before genet module to avoid the issue.

Fixes 9a4e79697009 ("net: bcmgenet: utilize generic Broadcom UniMAC MDIO controller driver")
Quoted 17 lines

---
v2: Load mdio-bcm-unimac before genet module instead of trying to
connect the PHY in a loop.

  drivers/net/ethernet/broadcom/genet/bcmgenet.c | 1 +
  1 file changed, 1 insertion(+)
drivers/net/ethernet/broadcom/genet/bcmgenet.c +1 -0
index fcca023f22e5..41f7f078cd27 100644
@@ -4296,3 +4296,4 @@ MODULE_AUTHOR("Broadcom Corporation");
  MODULE_DESCRIPTION("Broadcom GENET Ethernet controller driver");
  MODULE_ALIAS("platform:bcmgenet");
  MODULE_LICENSE("GPL");
+MODULE_SOFTDEP("pre: mdio-bcm-unimac");
Signature · 1 line
Florian
#12 lore raw
In reply to Jian-Hong Pan
This patch was applied to netdev/net.git (refs/heads/master):

Hello:

This patch was applied to netdev/net.git (refs/heads/master):

Quoted 13 lines · On Wed, 23 Jun 2021 11:28:03 +0800 you wrote:

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 1 line

[...]

Here is the summary with links:
- [v2] net: bcmgenet: Fix attaching to PYH failed on RPi 4B

    https://git.kernel.org/netdev/net/c/b2ac9800cfe0

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

#13 lore raw
Quoted 18 lines · On 6/23/21 1:50 PM, [email protected] wrote:

Hello:

This patch was applied to netdev/net.git (refs/heads/master):

Quoted 13 lines · On Wed, 23 Jun 2021 11:28:03 +0800 you wrote:

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 5 lines
Quoted 1 line

[...]

Here is the summary with links:
- [v2] net: bcmgenet: Fix attaching to PYH failed on RPi 4B

    https://git.kernel.org/netdev/net/c/b2ac9800cfe0

There was feedback given that could have deserved a v3, if nothing else to fix the typo in the subject, I suppose that would do though.

Signature · 1 line
Florian
#14 lore raw

Florian Fainelli <[email protected]> 於 2021年6月24日 週四 上午5:19寫道:

Quoted 20 lines · On 6/23/21 1:50 PM, [email protected] wrote:
Quoted 18 lines

Hello:

This patch was applied to netdev/net.git (refs/heads/master):

Quoted 13 lines · On Wed, 23 Jun 2021 11:28:03 +0800 you wrote:

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 5 lines
Quoted 5 lines
Quoted 1 line

[...]

Here is the summary with links:
- [v2] net: bcmgenet: Fix attaching to PYH failed on RPi 4B

    https://git.kernel.org/netdev/net/c/b2ac9800cfe0

This bot is interesting!!! Good feature! :)

Quoted 2 lines

There was feedback given that could have deserved a v3, if nothing else to fix the typo in the subject, I suppose that would do though.

I can prepare the v3 patch with Florian's suggestion!

Jian-Hong Pan

#15 lore raw
Quoted 27 lines · Jian-Hong Pan

Florian Fainelli <[email protected]> 於 2021年6月24日 週四 上午5:19寫道:

Quoted 20 lines · On 6/23/21 1:50 PM, [email protected] wrote:
Quoted 18 lines

Hello:

This patch was applied to netdev/net.git (refs/heads/master):

Quoted 13 lines · On Wed, 23 Jun 2021 11:28:03 +0800 you wrote:

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 5 lines
Quoted 5 lines
Quoted 1 line

[...]

Here is the summary with links:
- [v2] net: bcmgenet: Fix attaching to PYH failed on RPi 4B

     https://git.kernel.org/netdev/net/c/b2ac9800cfe0
Quoted 6 lines

This bot is interesting!!! Good feature! :)

Quoted 2 lines

There was feedback given that could have deserved a v3, if nothing else to fix the typo in the subject, I suppose that would do though.

I can prepare the v3 patch with Florian's suggestion!

Too late, once it's merged only fixups can be accepted, and that does not include commit messages, it's alright.

Signature · 1 line
Florian
#16 lore raw

Ok, thanks!

Florian Fainelli <[email protected]> 於 2021年6月24日 週四 上午11:27寫道:

Quoted 42 lines · Jian-Hong Pan
Quoted 27 lines

Florian Fainelli <[email protected]> 於 2021年6月24日 週四 上午5:19寫道:

Quoted 20 lines · On 6/23/21 1:50 PM, [email protected] wrote:
Quoted 18 lines

Hello:

This patch was applied to netdev/net.git (refs/heads/master):

Quoted 13 lines · On Wed, 23 Jun 2021 11:28:03 +0800 you wrote:

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 5 lines
Quoted 5 lines
Quoted 1 line

[...]

Here is the summary with links:
- [v2] net: bcmgenet: Fix attaching to PYH failed on RPi 4B

     https://git.kernel.org/netdev/net/c/b2ac9800cfe0
Quoted 6 lines

This bot is interesting!!! Good feature! :)

Quoted 2 lines

There was feedback given that could have deserved a v3, if nothing else to fix the typo in the subject, I suppose that would do though.

I can prepare the v3 patch with Florian's suggestion!

Too late, once it's merged only fixups can be accepted, and that does not include commit messages, it's alright.
--
Florian

#17 lore raw

PATCH v3 1/1 net: bcmgenet: Add mdio-bcm-unimac soft dependency

In reply to Florian Fainelli

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

It is not just coming too late, there is also no way for the module loader to figure out the dependency between GENET and its MDIO bus driver unless we provide this MODULE_SOFTDEP hint.

This patch adds the soft dependency to load mdio-bcm-unimac module before genet module to fix this issue.

Fixes 9a4e79697009 ("net: bcmgenet: utilize generic Broadcom UniMAC MDIO controller driver")

---
v3: Modified the subject and added some more description in the commit
message by Florian's suggestion.

v2: Load mdio-bcm-unimac before genet module instead of trying to
connect the PHY in a loop.

 drivers/net/ethernet/broadcom/genet/bcmgenet.c | 1 +
 1 file changed, 1 insertion(+)
drivers/net/ethernet/broadcom/genet/bcmgenet.c +1 -0
index fcca023f22e5..41f7f078cd27 100644
@@ -4296,3 +4296,4 @@ MODULE_AUTHOR("Broadcom Corporation");
 MODULE_DESCRIPTION("Broadcom GENET Ethernet controller driver");
 MODULE_ALIAS("platform:bcmgenet");
 MODULE_LICENSE("GPL");
+MODULE_SOFTDEP("pre: mdio-bcm-unimac");
Signature · 1 line
2.32.0
#18 lore raw

Re: [PATCH v3] net: bcmgenet: Add mdio-bcm-unimac soft dependency

This patch was applied to netdev/net-next.git (refs/heads/master):

Hello:

This patch was applied to netdev/net-next.git (refs/heads/master):

Quoted 13 lines · On Thu, 24 Jun 2021 11:22:41 +0800 you wrote:

The Broadcom UniMAC MDIO bus from mdio-bcm-unimac module comes too late. So, GENET cannot find the ethernet PHY on UniMAC MDIO bus. This leads GENET fail to attach the PHY as following log:

bcmgenet fd580000.ethernet: GENET 5.0 EPHY: 0x0000
...
could not attach to PHY
bcmgenet fd580000.ethernet eth0: failed to connect to PHY uart-pl011 fe201000.serial: no DMA platform data
libphy: bcmgenet MII bus: probed
...
unimac-mdio unimac-mdio.-19: Broadcom UniMAC MDIO bus

Quoted 1 line

[...]

Here is the summary with links:
- [v3] net: bcmgenet: Add mdio-bcm-unimac soft dependency

    https://git.kernel.org/netdev/net-next/c/19938bafa7ae

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html