[PATCH net-next v2] net: sfp: clean up i2c-bus property parsing

[PATCH net-next v2] net: sfp: clean up i2c-bus property parsing

[Russell King (Oracle)]

We currently have some complicated code in sfp_probe() which gets the
I2C bus depending on whether the sfp node is DT or ACPI, and we use
completely separate lookup functions.

This could do with being in a separate function to make the code more
readable, so move it to a new function, sfp_i2c_get(). We can also use
fwnode_find_reference() to lookup the I2C bus fwnode before then
decending into fwnode-type specific parsing.

A future cleanup would be to move the fwnode-type specific parsing into
the i2c layer, which is where it really should be.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
---
v2: actually send version with ACPI build error fixed (doesn't show up unless
 ACPI is enabled.)

 drivers/net/phy/sfp.c | 74 +++++++++++++++++++++++--------------------
 1 file changed, 40 insertions(+), 34 deletions(-)

diff --git a/drivers/net/phy/sfp.c b/drivers/net/phy/sfp.c
index 39fd1811375c..09c1d10f5b5d 100644
--- a/drivers/net/phy/sfp.c
+++ b/drivers/net/phy/sfp.c
@@ -2642,10 +2642,46 @@ static void sfp_cleanup(void *data)
 	kfree(sfp);
 }
 
+static int sfp_i2c_get(struct sfp *sfp)
+{
+	struct acpi_handle *acpi_handle;
+	struct fwnode_handle *h;
+	struct i2c_adapter *i2c;
+	struct device_node *np;
+	int err;
+
+	h = fwnode_find_reference(dev_fwnode(sfp->dev), "i2c-bus", 0);
+	if (IS_ERR(h)) {
+		dev_err(sfp->dev, "missing 'i2c-bus' property\n");
+		return -ENODEV;
+	}
+
+	if (is_acpi_device_node(h)) {
+		acpi_handle = ACPI_HANDLE_FWNODE(h);
+		i2c = i2c_acpi_find_adapter_by_handle(acpi_handle);
+	} else if ((np = to_of_node(h)) != NULL) {
+		i2c = of_find_i2c_adapter_by_node(np);
+	} else {
+		err = -EINVAL;
+		goto put;
+	}
+
+	if (!i2c) {
+		err = -EPROBE_DEFER;
+		goto put;
+	}
+
+	err = sfp_i2c_configure(sfp, i2c);
+	if (err)
+		i2c_put_adapter(i2c);
+put:
+	fwnode_handle_put(h);
+	return err;
+}
+
 static int sfp_probe(struct platform_device *pdev)
 {
 	const struct sff_data *sff;
-	struct i2c_adapter *i2c;
 	char *sfp_irq_name;
 	struct sfp *sfp;
 	int err, i;
@@ -2665,49 +2701,19 @@ static int sfp_probe(struct platform_device *pdev)
 	if (pdev->dev.of_node) {
 		struct device_node *node = pdev->dev.of_node;
 		const struct of_device_id *id;
-		struct device_node *np;
 
 		id = of_match_node(sfp_of_match, node);
 		if (WARN_ON(!id))
 			return -EINVAL;
 
 		sff = sfp->type = id->data;
-
-		np = of_parse_phandle(node, "i2c-bus", 0);
-		if (!np) {
-			dev_err(sfp->dev, "missing 'i2c-bus' property\n");
-			return -ENODEV;
-		}
-
-		i2c = of_find_i2c_adapter_by_node(np);
-		of_node_put(np);
-	} else if (has_acpi_companion(&pdev->dev)) {
-		struct acpi_device *adev = ACPI_COMPANION(&pdev->dev);
-		struct fwnode_handle *fw = acpi_fwnode_handle(adev);
-		struct fwnode_reference_args args;
-		struct acpi_handle *acpi_handle;
-		int ret;
-
-		ret = acpi_node_get_property_reference(fw, "i2c-bus", 0, &args);
-		if (ret || !is_acpi_device_node(args.fwnode)) {
-			dev_err(&pdev->dev, "missing 'i2c-bus' property\n");
-			return -ENODEV;
-		}
-
-		acpi_handle = ACPI_HANDLE_FWNODE(args.fwnode);
-		i2c = i2c_acpi_find_adapter_by_handle(acpi_handle);
-	} else {
+	} else if (!has_acpi_companion(&pdev->dev)) {
 		return -EINVAL;
 	}
 
-	if (!i2c)
-		return -EPROBE_DEFER;
-
-	err = sfp_i2c_configure(sfp, i2c);
-	if (err < 0) {
-		i2c_put_adapter(i2c);
+	err = sfp_i2c_get(sfp);
+	if (err)
 		return err;
-	}
 
 	for (i = 0; i < GPIO_MAX; i++)
 		if (sff->gpios & BIT(i)) {
-- 
2.30.2

Re: [PATCH net-next v2] net: sfp: clean up i2c-bus property parsing

[Russell King (Oracle)]

On Sun, Dec 04, 2022 at 12:40:25AM +0800, kernel test robot wrote:
> Thank you for the patch! Perhaps something to improve:

Sigh... this is another stupidity.

> 73970055450eeb Russell King          2017-07-25  2701  	if (pdev->dev.of_node) {
> 73970055450eeb Russell King          2017-07-25 @2702  		struct device_node *node = pdev->dev.of_node;

"node" declared here...

> 259c8618b0099b Russell King          2017-12-14  2703  		const struct of_device_id *id;
> 73970055450eeb Russell King          2017-07-25  2704  
> 259c8618b0099b Russell King          2017-12-14  2705  		id = of_match_node(sfp_of_match, node);

... and clearly used here, so the code looks to be correct.

However, when CONFIG_OF is not set, of_match_node() does not make use
of this argument:

#define of_match_node(_matches, _node)  NULL

which results in otherwise correct code issuing a warning when
CONFIG_OF is disabled... and sure enough, your configuration has:

> # CONFIG_OF is not set

This illustrates just how bad an idea it is to use compiler macros for
this stuff - it actively hurts compile testing, because you have to
test every damn combination of configuration options to get proper
coverage, which is totally and utterly rediculous.

of_match_node() and ACPI_HANDLE_FWNODE() should *both* be inline
functions when the subsystem is disabled, so that incorrect arguments
can be detected, and warnings about unused variables such as the one
you're reporting here doesn't happen.

While the issue lies firmly in the realms of the DT (and ACPI) headers,
I will yet again respin this patch to sort this out - but really the
correct solution is to fix the bloody headers so compile coverage
actually works.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!