[PATCH v2 1/2] phy dp83867: Fix compilation with CONFIG_OF_MDIO=m

Subsystems: ethernet phy library, networking drivers, the rest

STALE3733d

12 messages, 5 authors, 2016-05-17 · open the first message on its own page

[PATCH v2 1/2] phy dp83867: Fix compilation with CONFIG_OF_MDIO=m

From: Alexander Graf <hidden>
Date: 2016-05-16 18:52:46

When CONFIG_OF_MDIO is configured as module, the #define for it really
is CONFIG_OF_MDIO_MODULE, not CONFIG_OF_MDIO. So if we are compiling it
as module, the dp83867 doesn't see that OF_MDIO was selected and doesn't
read the dt rgmii parameters.

The fix is simple: Use IS_ENABLED(). It checks for both - module as well
as compiled in code.

Signed-off-by: Alexander Graf <redacted>
---
 drivers/net/phy/dp83867.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/phy/dp83867.c b/drivers/net/phy/dp83867.c
index 2afa61b..94cc278 100644
--- a/drivers/net/phy/dp83867.c
+++ b/drivers/net/phy/dp83867.c

@@ -99,7 +99,7 @@ static int dp83867_config_intr(struct phy_device *phydev)
 	return phy_write(phydev, MII_DP83867_MICR, micr_status);
 }
 
-#ifdef CONFIG_OF_MDIO
+#if IS_ENABLED(CONFIG_OF_MDIO)
 static int dp83867_of_init(struct phy_device *phydev)
 {
 	struct dp83867_private *dp83867 = phydev->priv;

-- 
1.8.5.6

[PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Alexander Graf <hidden>
Date: 2016-05-16 18:52:48

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>
---
 drivers/net/phy/dp83867.c | 31 ++++++++++++++++++++++++++++---
 1 file changed, 28 insertions(+), 3 deletions(-)

diff --git a/drivers/net/phy/dp83867.c b/drivers/net/phy/dp83867.c
index 94cc278..1b01680 100644
--- a/drivers/net/phy/dp83867.c
+++ b/drivers/net/phy/dp83867.c

@@ -65,6 +65,7 @@ struct dp83867_private {
 	int rx_id_delay;
 	int tx_id_delay;
 	int fifo_depth;
+	int values_are_sane;
 };
 
 static int dp83867_ack_interrupt(struct phy_device *phydev)

@@ -113,15 +114,30 @@ static int dp83867_of_init(struct phy_device *phydev)
 	ret = of_property_read_u32(of_node, "ti,rx-internal-delay",
 				   &dp83867->rx_id_delay);
 	if (ret)
-		return ret;
+		goto invalid_dt;
 
 	ret = of_property_read_u32(of_node, "ti,tx-internal-delay",
 				   &dp83867->tx_id_delay);
 	if (ret)
-		return ret;
+		goto invalid_dt;
 
-	return of_property_read_u32(of_node, "ti,fifo-depth",
+	ret = of_property_read_u32(of_node, "ti,fifo-depth",
 				   &dp83867->fifo_depth);
+	if (ret)
+		goto invalid_dt;
+
+	dp83867->values_are_sane = 1;
+
+	return 0;
+
+invalid_dt:
+	phydev_err(phydev, "missing properties in device tree");
+
+	/*
+	 * We can still run with a broken dt by not using any of the optional
+	 * parameters, so just don't set dp83867->values_are_sane.
+	 */
+	return 0;
 }
 #else
 static int dp83867_of_init(struct phy_device *phydev)

@@ -150,6 +166,15 @@ static int dp83867_config_init(struct phy_device *phydev)
 		dp83867 = (struct dp83867_private *)phydev->priv;
 	}
 
+	/*
+	 * With no or broken device tree, we don't have the values that we would
+	 * want to configure the phy with. In that case, cross our fingers and
+	 * assume that firmware did everything correctly for us or that we don't
+	 * need them.
+	 */
+	if (!dp83867->values_are_sane)
+		return 0;
+
 	if (phy_interface_is_rgmii(phydev)) {
 		ret = phy_write(phydev, MII_DP83867_PHYCTRL,
 			(dp83867->fifo_depth << DP83867_PHYCR_FIFO_DEPTH_SHIFT));

-- 
1.8.5.6

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Florian Fainelli <f.fainelli@gmail.com>
Date: 2016-05-16 19:04:10

On 05/16/2016 11:52 AM, Alexander Graf wrote:

quoted hunk

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>
---
 drivers/net/phy/dp83867.c | 31 ++++++++++++++++++++++++++++---
 1 file changed, 28 insertions(+), 3 deletions(-)

diff --git a/drivers/net/phy/dp83867.c b/drivers/net/phy/dp83867.c
index 94cc278..1b01680 100644
--- a/drivers/net/phy/dp83867.c
+++ b/drivers/net/phy/dp83867.c

@@ -65,6 +65,7 @@ struct dp83867_private {
 	int rx_id_delay;
 	int tx_id_delay;
 	int fifo_depth;
+	int values_are_sane;

This could be a boolean type.

quoted hunk

 };
 
 static int dp83867_ack_interrupt(struct phy_device *phydev)

@@ -113,15 +114,30 @@ static int dp83867_of_init(struct phy_device *phydev)
 	ret = of_property_read_u32(of_node, "ti,rx-internal-delay",
 				   &dp83867->rx_id_delay);
 	if (ret)
-		return ret;
+		goto invalid_dt;
 
 	ret = of_property_read_u32(of_node, "ti,tx-internal-delay",
 				   &dp83867->tx_id_delay);
 	if (ret)
-		return ret;
+		goto invalid_dt;
 
-	return of_property_read_u32(of_node, "ti,fifo-depth",
+	ret = of_property_read_u32(of_node, "ti,fifo-depth",
 				   &dp83867->fifo_depth);
+	if (ret)
+		goto invalid_dt;
+
+	dp83867->values_are_sane = 1;
+
+	return 0;
+
+invalid_dt:
+	phydev_err(phydev, "missing properties in device tree");

phydev_warn() maybe?

Other than that, this looks okay to me.
-- 
Florian

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Andrew Lunn <andrew@lunn.ch>
Date: 2016-05-16 20:15:15

On Mon, May 16, 2016 at 08:52:43PM +0200, Alexander Graf wrote:

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

The binding document actually says they are required. It would be good
to make the binding documentation and the code consistent.

   Andrew

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Dan Murphy <hidden>
Date: 2016-05-16 20:28:03

On 05/16/2016 01:52 PM, Alexander Graf wrote:

quoted hunk

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>
---
 drivers/net/phy/dp83867.c | 31 ++++++++++++++++++++++++++++---
 1 file changed, 28 insertions(+), 3 deletions(-)

diff --git a/drivers/net/phy/dp83867.c b/drivers/net/phy/dp83867.c
index 94cc278..1b01680 100644
--- a/drivers/net/phy/dp83867.c
+++ b/drivers/net/phy/dp83867.c

@@ -65,6 +65,7 @@ struct dp83867_private {
 	int rx_id_delay;
 	int tx_id_delay;
 	int fifo_depth;
+	int values_are_sane;
 };
 
 static int dp83867_ack_interrupt(struct phy_device *phydev)

@@ -113,15 +114,30 @@ static int dp83867_of_init(struct phy_device *phydev)
 	ret = of_property_read_u32(of_node, "ti,rx-internal-delay",
 				   &dp83867->rx_id_delay);
 	if (ret)
-		return ret;
+		goto invalid_dt;
 
 	ret = of_property_read_u32(of_node, "ti,tx-internal-delay",
 				   &dp83867->tx_id_delay);
 	if (ret)
-		return ret;
+		goto invalid_dt;

Optional means you may or may not have the entries

I would prefer to wrap the DT reading with the interface type check.

if (phydev->interface == PHY_INTERFACE_MODE_RGMII_TXID ||
phydev->interface == PHY_INTERFACE_MODE_RGMII_ID  )

 	ret = of_property_read_u32(of_node, "ti,tx-internal-delay",
 				   &dp83867->tx_id_delay);
	if (ret)

goto invalid_dt;

Otherwise this continues to mandate that you need to declare all the DT entries
when in fact you may only have to declare 1.

And if the other interfaces are declared then DT entries are ignored.  And configuring
internal delay is not required per section 8.9 footnote 3 of the data sheet.

Dan

quoted hunk

 
-	return of_property_read_u32(of_node, "ti,fifo-depth",
+	ret = of_property_read_u32(of_node, "ti,fifo-depth",
 				   &dp83867->fifo_depth);
+	if (ret)
+		goto invalid_dt;
+
+	dp83867->values_are_sane = 1;
+
+	return 0;
+
+invalid_dt:
+	phydev_err(phydev, "missing properties in device tree");
+
+	/*
+	 * We can still run with a broken dt by not using any of the optional
+	 * parameters, so just don't set dp83867->values_are_sane.
+	 */
+	return 0;
 }
 #else
 static int dp83867_of_init(struct phy_device *phydev)

@@ -150,6 +166,15 @@ static int dp83867_config_init(struct phy_device *phydev)
 		dp83867 = (struct dp83867_private *)phydev->priv;
 	}
 
+	/*
+	 * With no or broken device tree, we don't have the values that we would
+	 * want to configure the phy with. In that case, cross our fingers and
+	 * assume that firmware did everything correctly for us or that we don't
+	 * need them.
+	 */
+	if (!dp83867->values_are_sane)
+		return 0;
+
 	if (phy_interface_is_rgmii(phydev)) {
 		ret = phy_write(phydev, MII_DP83867_PHYCTRL,
 			(dp83867->fifo_depth << DP83867_PHYCR_FIFO_DEPTH_SHIFT));


-- 
------------------
Dan Murphy

Re: [PATCH v2 1/2] phy dp83867: Fix compilation with CONFIG_OF_MDIO=m

From: David Miller <davem@davemloft.net>
Date: 2016-05-17 18:22:23

From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:42 +0200

When CONFIG_OF_MDIO is configured as module, the #define for it really
is CONFIG_OF_MDIO_MODULE, not CONFIG_OF_MDIO. So if we are compiling it
as module, the dp83867 doesn't see that OF_MDIO was selected and doesn't
read the dt rgmii parameters.

The fix is simple: Use IS_ENABLED(). It checks for both - module as well
as compiled in code.

Signed-off-by: Alexander Graf <redacted>

Applied.

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: David Miller <davem@davemloft.net>
Date: 2016-05-17 18:22:28

From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:43 +0200

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>

Applied.

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Dan Murphy <hidden>
Date: 2016-05-17 18:34:47

David

On 05/17/2016 01:22 PM, David Miller wrote:

From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:43 +0200

quoted

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>

Applied.

This patch should not have been applied.

I did not believe the implementation was proper for that driver.

It seems my objection to the code was not seen.  Nor was Andrew's point about the DT bindings document

https://patchwork.kernel.org/patch/9105371/

Dan

-- 
------------------
Dan Murphy

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: David Miller <davem@davemloft.net>
Date: 2016-05-17 18:48:42

From: Dan Murphy <redacted>
Date: Tue, 17 May 2016 13:34:34 -0500

David

On 05/17/2016 01:22 PM, David Miller wrote:

quoted

From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:43 +0200

quoted

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>

Applied.

This patch should not have been applied.

I did not believe the implementation was proper for that driver.

It seems my objection to the code was not seen.  Nor was Andrew's point about the DT bindings document

https://patchwork.kernel.org/patch/9105371/

The discussions around the recent phy patches have been a labrynth that I've
found hard to follow, sorry.

I'll revert these two, sigh....

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Alexander Graf <hidden>
Date: 2016-05-17 20:37:40

Hi David,

Am 17.05.2016 um 20:48 schrieb David Miller [off-list ref]:

From: Dan Murphy <redacted>
Date: Tue, 17 May 2016 13:34:34 -0500

quoted

David

quoted

On 05/17/2016 01:22 PM, David Miller wrote:
From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:43 +0200

quoted

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>

Applied.

This patch should not have been applied.

I did not believe the implementation was proper for that driver.

It seems my objection to the code was not seen.  Nor was Andrew's point about the DT bindings document

https://patchwork.kernel.org/patch/9105371/

The discussions around the recent phy patches have been a labrynth that I've
found hard to follow, sorry.

I'll revert these two, sigh....

The first patch is an obvious and correct fix. Discussions were only about the second one (which I'm happy to drop given the rat hole this turned out to be).

Alex

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Dan Murphy <hidden>
Date: 2016-05-17 20:40:20

On 05/17/2016 03:37 PM, Alexander Graf wrote:

Hi David,

quoted

Am 17.05.2016 um 20:48 schrieb David Miller [off-list ref]:

From: Dan Murphy <redacted>
Date: Tue, 17 May 2016 13:34:34 -0500

quoted

David

quoted

On 05/17/2016 01:22 PM, David Miller wrote:
From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:43 +0200

quoted

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>

Applied.

This patch should not have been applied.

I did not believe the implementation was proper for that driver.

It seems my objection to the code was not seen.  Nor was Andrew's point about the DT bindings document

https://patchwork.kernel.org/patch/9105371/

The discussions around the recent phy patches have been a labrynth that I've
found hard to follow, sorry.

I'll revert these two, sigh....

The first patch is an obvious and correct fix. Discussions were only about the second one (which I'm happy to drop given the rat hole this turned out to be).

So are you going to abandon the second patch all together?
If you do, let me know I can submit a patch.

Dan

Alex


-- 
------------------
Dan Murphy

Re: [PATCH v2 2/2] phy dp83867: Make rgmii parameters optional

From: Alexander Graf <hidden>
Date: 2016-05-17 20:45:45

Am 17.05.2016 um 22:40 schrieb Dan Murphy [off-list ref]:

quoted

On 05/17/2016 03:37 PM, Alexander Graf wrote:
Hi David,

quoted

Am 17.05.2016 um 20:48 schrieb David Miller [off-list ref]:

From: Dan Murphy <redacted>
Date: Tue, 17 May 2016 13:34:34 -0500

quoted

David

quoted

On 05/17/2016 01:22 PM, David Miller wrote:
From: Alexander Graf <redacted>
Date: Mon, 16 May 2016 20:52:43 +0200

quoted

If you compile without OF_MDIO support in an RGMII configuration, we fail
to configure the dp83867 phy today by writing garbage into its configuration
registers.

On the other hand if you do compile with OF_MDIO and the phy gets loaded via
device tree, you have to have the properties set in the device tree, otherwise
we fail to load the driver and don't even attach the generic phy driver to
the interface anymore.

To make things slightly more consistent, make the rgmii configuration properties
optional and allow a user to omit them in their device tree.

Signed-off-by: Alexander Graf <redacted>

Applied.

This patch should not have been applied.

I did not believe the implementation was proper for that driver.

It seems my objection to the code was not seen.  Nor was Andrew's point about the DT bindings document

https://patchwork.kernel.org/patch/9105371/

The discussions around the recent phy patches have been a labrynth that I've
found hard to follow, sorry.

I'll revert these two, sigh....

The first patch is an obvious and correct fix. Discussions were only about the second one (which I'm happy to drop given the rat hole this turned out to be).

So are you going to abandon the second patch all together?
If you do, let me know I can submit a patch.

I'm fairly sure you'll be done with it much quicker than me and the real bug I wanted to fix is that the driver broke with OF_MDIO=m ;)

So yes, please take over from here.


Alex

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help