Skip to contentSkip to conversation

Revert "net: stmmac: platform: Fix MDIO init for platforms without PHY"

PATCH NAK Florian Fainelli 5 participants 7 messages Jan 7, 2020

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

PATCH 1/1 Revert "net: stmmac: platform: Fix MDIO init for platforms without PHY"

This reverts commit d3e014ec7d5ebe9644b5486bc530b91e62bbf624 ("net:
stmmac: platform: Fix MDIO init for platforms without PHY") because it breaks existing systems with stmmac which do not have a MDIO bus sub-node nor a 'phy-handle' property declared in their Device Tree. On those systems, the stmmac MDIO bus is expected to be created and then scanned by of_mdiobus_register() to create PHY devices.

While these systems should arguably make use of a more accurate Device Tree reprensentation with the use of the MDIO bus sub-node an appropriate 'phy-handle', we cannot break them, therefore restore the behavior prior to the said commit.

Fixes d3e014ec7d5e ("net: stmmac: platform: Fix MDIO init for platforms without PHY")

---
Heiko,

I did not add the Tested-by because the patch is a little bit different from what you tested, even if you most likely were not hitting the other part that I was changing. Thanks!

 drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +1 -1
index cc8d7e7bf9ac..bedaff0c13bd 100644
@@ -320,7 +320,7 @@ static int stmmac_mtl_setup(struct platform_device *pdev,
 static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
 			 struct device_node *np, struct device *dev)
 {
-	bool mdio = false;
+	bool mdio = true;
 	static const struct of_device_id need_mdio_ids[] = {
 		{ .compatible = "snps,dwc-qos-ethernet-4.10" },
 		{},
Signature · 1 line
2.19.1
Nacked-by on v1 1/1 NAK
#1 lore raw
Quoted 2 lines

From: Florian Fainelli <[email protected]>
Subject: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for

platforms

Quoted 4 lines

without PHY"

This reverts commit d3e014ec7d5ebe9644b5486bc530b91e62bbf624 ("net:
stmmac: platform: Fix MDIO init for platforms without PHY") because it

breaks

Quoted 6 lines

existing systems with stmmac which do not have a MDIO bus sub-node nor a 'phy-handle' property declared in their Device Tree. On those systems, the stmmac MDIO bus is expected to be created and then scanned by of_mdiobus_register() to create PHY devices.

While these systems should arguably make use of a more accurate Device

Tree

Quoted 2 lines

reprensentation with the use of the MDIO bus sub-node an appropriate 'phy- handle', we cannot break them, therefore restore the behavior prior to the

said

Quoted 7 lines

commit.

Fixes d3e014ec7d5e ("net: stmmac: platform: Fix MDIO init for platforms

without PHY")

Quoted 4 lines

---
Heiko,

I did not add the Tested-by because the patch is a little bit different

from what

Quoted 1 line

you tested, even if you most likely were not hitting the other part that I

was

Quoted 11 lines

changing. Thanks!

 drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -0
index cc8d7e7bf9ac..bedaff0c13bd 100644

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c static int stmmac_mtl_setup(struct platform_device · line 322

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +1 -1
@@ -320,7 +320,7 @@ static int stmmac_mtl_setup(struct platform_device
 *pdev,  static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
 			 struct device_node *np, struct device *dev)  {
-	bool mdio = false;
+	bool mdio = true;

This is breaking for the platforms with fixed-link. stih418-b2199.dts and 169445.dts to name a few.

For the newer platforms, they should provide the mdio/ snps,dwmac-mdio property in the device tree as we are checking the mdio/ snps,dwmac-mdio property in the stmmac_platform driver for the mdio bus memory allocation. For existing platforms, I agree we should not break them, but we should make the code correct. And make the existing platforms adapt to the proper code. There is a proposed solution.
https://lkml.org/lkml/2020/1/7/14

What do you think?

Quoted 5 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -1
 	static const struct of_device_id need_mdio_ids[] = {
 		{ .compatible = "snps,dwc-qos-ethernet-4.10" },
 		{},
--

2.19.1

#2 lore raw
Quoted 54 lines · Sriram Dash
Quoted 2 lines

From: Florian Fainelli <[email protected]>
Subject: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for

platforms

Quoted 4 lines

without PHY"

This reverts commit d3e014ec7d5ebe9644b5486bc530b91e62bbf624 ("net:
stmmac: platform: Fix MDIO init for platforms without PHY") because it

breaks

Quoted 6 lines

existing systems with stmmac which do not have a MDIO bus sub-node nor a 'phy-handle' property declared in their Device Tree. On those systems, the stmmac MDIO bus is expected to be created and then scanned by of_mdiobus_register() to create PHY devices.

While these systems should arguably make use of a more accurate Device

Tree

Quoted 2 lines

reprensentation with the use of the MDIO bus sub-node an appropriate 'phy- handle', we cannot break them, therefore restore the behavior prior to the

said

Quoted 7 lines

commit.

Fixes d3e014ec7d5e ("net: stmmac: platform: Fix MDIO init for platforms

without PHY")

Quoted 4 lines

---
Heiko,

I did not add the Tested-by because the patch is a little bit different

from what

Quoted 1 line

you tested, even if you most likely were not hitting the other part that I

was

Quoted 11 lines

changing. Thanks!

 drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -0
index cc8d7e7bf9ac..bedaff0c13bd 100644

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c static int stmmac_mtl_setup(struct platform_device · line 322

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +1 -1
@@ -320,7 +320,7 @@ static int stmmac_mtl_setup(struct platform_device
 *pdev,  static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
 			 struct device_node *np, struct device *dev)  {
-	bool mdio = false;
+	bool mdio = true;

This is breaking for the platforms with fixed-link. stih418-b2199.dts and 169445.dts to name a few.

Quoted 7 lines

For the newer platforms, they should provide the mdio/ snps,dwmac-mdio property in the device tree as we are checking the mdio/ snps,dwmac-mdio property in the stmmac_platform driver for the mdio bus memory allocation. For existing platforms, I agree we should not break them, but we should make the code correct. And make the existing platforms adapt to the proper code. There is a proposed solution.
https://lkml.org/lkml/2020/1/7/14

Can you post it as a patch then and add me in CC since it also breaks Oxnas as reported as Kci ?

Neil

Quoted 16 lines

What do you think?

Quoted 5 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -1
 	static const struct of_device_id need_mdio_ids[] = {
 		{ .compatible = "snps,dwc-qos-ethernet-4.10" },
 		{},
--

2.19.1

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

#3 lore raw
In reply to Sriram Dash
Quoted 62 lines · Sriram Dash
Quoted 2 lines

From: Florian Fainelli <[email protected]>
Subject: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for

platforms

Quoted 4 lines

without PHY"

This reverts commit d3e014ec7d5ebe9644b5486bc530b91e62bbf624 ("net:
stmmac: platform: Fix MDIO init for platforms without PHY") because it

breaks

Quoted 6 lines

existing systems with stmmac which do not have a MDIO bus sub-node nor a 'phy-handle' property declared in their Device Tree. On those systems, the stmmac MDIO bus is expected to be created and then scanned by of_mdiobus_register() to create PHY devices.

While these systems should arguably make use of a more accurate Device

Tree

Quoted 2 lines

reprensentation with the use of the MDIO bus sub-node an appropriate 'phy- handle', we cannot break them, therefore restore the behavior prior to the

said

Quoted 7 lines

commit.

Fixes d3e014ec7d5e ("net: stmmac: platform: Fix MDIO init for platforms

without PHY")

Quoted 4 lines

---
Heiko,

I did not add the Tested-by because the patch is a little bit different

from what

Quoted 1 line

you tested, even if you most likely were not hitting the other part that I

was

Quoted 11 lines

changing. Thanks!

  drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -0
index cc8d7e7bf9ac..bedaff0c13bd 100644

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c static int stmmac_mtl_setup(struct platform_device · line 322

Quoted 4 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +1 -1
@@ -320,7 +320,7 @@ static int stmmac_mtl_setup(struct platform_device
 *pdev,  static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
  			 struct device_node *np, struct device *dev)  {
-	bool mdio = false;
+	bool mdio = true;

This is breaking for the platforms with fixed-link. stih418-b2199.dts and 169445.dts to name a few.

For the newer platforms, they should provide the mdio/ snps,dwmac-mdio property in the device tree as we are checking the mdio/ snps,dwmac-mdio property in the stmmac_platform driver for the mdio bus memory allocation. For existing platforms, I agree we should not break them, but we should make the code correct. And make the existing platforms adapt to the proper code. There is a proposed solution.
https://lkml.org/lkml/2020/1/7/14

Quoted 1 line

What do you think?

The binding says that the phy handle and mdio child node are optional, so "update all of the DTBs!" is not a viable solution. I'm far from an expert here, but AFAICS the fault of the current code is that it assumes the lack of a phy handle implies a fixed link, so the obvious answer is to actually check whether the "fixed-link" property is present.

Robin.

Quoted 14 lines
Quoted 6 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -1
  	static const struct of_device_id need_mdio_ids[] = {
  		{ .compatible = "snps,dwc-qos-ethernet-4.10" },
  		{},
--

2.19.1

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

#4 lore raw
Quoted 73 lines

From: Robin Murphy <[email protected]>
Subject: Re: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for platforms without PHY"

Quoted 45 lines · Sriram Dash
Quoted 2 lines

From: Florian Fainelli <[email protected]>
Subject: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for

platforms

Quoted 5 lines

without PHY"

This reverts commit d3e014ec7d5ebe9644b5486bc530b91e62bbf624 ("net:
stmmac: platform: Fix MDIO init for platforms without PHY") because it

breaks

Quoted 8 lines

existing systems with stmmac which do not have a MDIO bus sub-node nor a 'phy-handle' property declared in their Device Tree. On those systems, the stmmac MDIO bus is expected to be created and then scanned by
of_mdiobus_register() to create PHY devices.

While these systems should arguably make use of a more accurate Device

Tree

Quoted 3 lines

reprensentation with the use of the MDIO bus sub-node an appropriate 'phy- handle', we cannot break them, therefore restore the behavior prior to the

said

Quoted 7 lines

commit.

Fixes d3e014ec7d5e ("net: stmmac: platform: Fix MDIO init for

platforms without PHY")

Quoted 5 lines

---
Heiko,

I did not add the Tested-by because the patch is a little bit different

from what

Quoted 2 lines

you tested, even if you most likely were not hitting the other part that I

was

Quoted 5 lines

changing. Thanks!

  drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 7 lines
Quoted 7 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -0
index cc8d7e7bf9ac..bedaff0c13bd 100644
@@ -320,7 +320,7 @@ static int stmmac_mtl_setup(struct

platform_device *pdev, static int stmmac_dt_phy(struct

plat_stmmacenet_data *plat,

Quoted 11 lines

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c

Quoted 3 lines
  			 struct device_node *np, struct device *dev)  {
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +1 -1
-	bool mdio = false;
+	bool mdio = true;

This is breaking for the platforms with fixed-link. stih418-b2199.dts and 169445.dts to name a few.

For the newer platforms, they should provide the mdio/ snps,dwmac-mdio property in the device tree as we are checking the mdio/ snps,dwmac-mdio property in the stmmac_platform driver for the mdio bus

memory allocation.

Quoted 2 lines

For existing platforms, I agree we should not break them, but we should make the code correct. And make the existing platforms adapt to the

proper code.

Quoted 8 lines
Quoted 5 lines

There is a proposed solution.
https://protect2.fireeye.com/url?k=d075c0fc-8da69942-d0744bb3-0cc47a31
ba82-60be9f5a0fb38a27&u=https://lkml.org/lkml/2020/1/7/14

What do you think?

The binding says that the phy handle and mdio child node are optional, so "update all of the DTBs!" is not a viable solution. I'm far from an expert here, but

I get your point Robin. Because the mdio child or snps,dwmac-mdio is not mandatory, there has been this problem.

Quoted 2 lines

AFAICS the fault of the current code is that it assumes the lack of a phy handle implies a fixed link, so the obvious answer is to actually check whether the

Kind of.

Quoted 1 line

"fixed-link" property is present.

I agree to have a fixed link check. Possibly this can be done:

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +5 -4
@@ -320,7 +320,9 @@ static int stmmac_mtl_setup(struct platform_device *pdev,
 static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
                         struct device_node *np, struct device *dev)
 {
-       bool mdio = false;
+       bool mdio = true;
+       struct device_node *fixed_link;
+
        static const struct of_device_id need_mdio_ids[] = {
                { .compatible = "snps,dwc-qos-ethernet-4.10" },
                {},
@@ -340,9 +342,8 @@ static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
                }
        }
 
-       if (plat->mdio_node) {
-               dev_dbg(dev, "Found MDIO subnode\n");
-               mdio = true;
+       if (of_get_child_by_name(np, "fixed-link")) {
+               mdio = false;
        }
 
        if (mdio) {

Neil and Heiko,

Can you guys please test this on your platforms?
We can post a more cleaner version of the patch if all agree to it.

Quoted 19 lines

Robin.

Quoted 16 lines
Quoted 6 lines

static const struct of_device_id need_mdio_ids[] = {
{ .compatible = "snps,dwc-qos-ethernet-4.10" }, {},
--
2.19.1

_______________________________________________
linux-arm-kernel mailing list [email protected]
https://protect2.fireeye.com/url?k=2cb95551-716a0cef-2cb8de1e-0cc47a31
ba82-d01ceb62a8a93fa2&u=http://lists.infradead.org/mailman/listinfo/li nux-arm-kernel

#5 lore raw
Quoted 88 lines · Sriram Dash
Quoted 73 lines

From: Robin Murphy <[email protected]>
Subject: Re: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for platforms without PHY"

Quoted 45 lines · Sriram Dash
Quoted 2 lines

From: Florian Fainelli <[email protected]>
Subject: [PATCH net] Revert "net: stmmac: platform: Fix MDIO init for

platforms

Quoted 5 lines

without PHY"

This reverts commit d3e014ec7d5ebe9644b5486bc530b91e62bbf624 ("net:
stmmac: platform: Fix MDIO init for platforms without PHY") because it

breaks

Quoted 8 lines

existing systems with stmmac which do not have a MDIO bus sub-node nor a 'phy-handle' property declared in their Device Tree. On those systems, the stmmac MDIO bus is expected to be created and then scanned by
of_mdiobus_register() to create PHY devices.

While these systems should arguably make use of a more accurate Device

Tree

Quoted 3 lines

reprensentation with the use of the MDIO bus sub-node an appropriate 'phy- handle', we cannot break them, therefore restore the behavior prior to the

said

Quoted 7 lines

commit.

Fixes d3e014ec7d5e ("net: stmmac: platform: Fix MDIO init for

platforms without PHY")

Quoted 5 lines

---
Heiko,

I did not add the Tested-by because the patch is a little bit different

from what

Quoted 2 lines

you tested, even if you most likely were not hitting the other part that I

was

Quoted 5 lines

changing. Thanks!

   drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 7 lines
Quoted 7 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +0 -0
index cc8d7e7bf9ac..bedaff0c13bd 100644
@@ -320,7 +320,7 @@ static int stmmac_mtl_setup(struct

platform_device *pdev, static int stmmac_dt_phy(struct

plat_stmmacenet_data *plat,

Quoted 11 lines

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c static int stmmac_mtl_setup(struct platform_device *pdev, · line 323

Quoted 3 lines
   			 struct device_node *np, struct device *dev)  {
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +1 -1
-	bool mdio = false;
+	bool mdio = true;

This is breaking for the platforms with fixed-link. stih418-b2199.dts and 169445.dts to name a few.

For the newer platforms, they should provide the mdio/ snps,dwmac-mdio property in the device tree as we are checking the mdio/ snps,dwmac-mdio property in the stmmac_platform driver for the mdio bus

memory allocation.

Quoted 2 lines

For existing platforms, I agree we should not break them, but we should make the code correct. And make the existing platforms adapt to the

proper code.

Quoted 8 lines
Quoted 5 lines

There is a proposed solution.
https://protect2.fireeye.com/url?k=d075c0fc-8da69942-d0744bb3-0cc47a31
ba82-60be9f5a0fb38a27&u=https://lkml.org/lkml/2020/1/7/14

What do you think?

The binding says that the phy handle and mdio child node are optional, so "update all of the DTBs!" is not a viable solution. I'm far from an expert here, but

I get your point Robin. Because the mdio child or snps,dwmac-mdio is not mandatory, there has been this problem.

Quoted 2 lines

AFAICS the fault of the current code is that it assumes the lack of a phy handle implies a fixed link, so the obvious answer is to actually check whether the

Quoted 6 lines

Kind of.

Quoted 1 line

"fixed-link" property is present.

I agree to have a fixed link check. Possibly this can be done:

FWIW, given the 4th case in the kerneldoc table I think this check might need to come first. On the upside, there appears to be a handy helper already, so it may just be as simple as:

	bool mdio = !of_phy_is_fixed_link(np);

Robin.

Quoted 51 lines
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c +5 -4
@@ -320,7 +320,9 @@ static int stmmac_mtl_setup(struct platform_device *pdev,
  static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
                          struct device_node *np, struct device *dev)
  {
-       bool mdio = false;
+       bool mdio = true;
+       struct device_node *fixed_link;
+
         static const struct of_device_id need_mdio_ids[] = {
                 { .compatible = "snps,dwc-qos-ethernet-4.10" },
                 {},
@@ -340,9 +342,8 @@ static int stmmac_dt_phy(struct plat_stmmacenet_data *plat,
                 }
         }
 
-       if (plat->mdio_node) {
-               dev_dbg(dev, "Found MDIO subnode\n");
-               mdio = true;
+       if (of_get_child_by_name(np, "fixed-link")) {
+               mdio = false;
         }
 
         if (mdio) {

Neil and Heiko,

Can you guys please test this on your platforms?
We can post a more cleaner version of the patch if all agree to it.

Quoted 19 lines

Robin.

Quoted 16 lines
Quoted 6 lines
   	static const struct of_device_id need_mdio_ids[] = {
   		{ .compatible = "snps,dwc-qos-ethernet-4.10" },
   		{},

--
2.19.1

_______________________________________________
linux-arm-kernel mailing list [email protected]
https://protect2.fireeye.com/url?k=2cb95551-716a0cef-2cb8de1e-0cc47a31
ba82-d01ceb62a8a93fa2&u=http://lists.infradead.org/mailman/listinfo/li nux-arm-kernel