Skip to contentSkip to conversation

Improve error handling in Rockchip Inno USB 2.0 PHY driver

PATCH Superseded Dragan Simic 3 participants 14 messages Aug 21, 2024

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover Improve error handling in Rockchip Inno USB 2.0 PHY driver

This is a small series that improves error handling in the probe path of the Rockchip Innosilicon USB 2.0 PHY driver, by returning the actual error code in one place and by using dev_err_probe() properly in multiple places. It also performs a bunch of small, rather trivial code cleanups, to make the code neater and a bit easier to read.

Changes in v2:
- Expanded into a small series, after a suggestion from Heiko [1] to use dev_err_probe(), because it all happens in the probe path

Link to v1: https://lore.kernel.org/linux-phy/5fa7796d71e2f46344e972bc98a54539f55b6109.1723551599.git.dsimic@manjaro.org/T/#u

[1] https://lore.kernel.org/linux-phy/4927264.xgNZFEDtJV@diego/

Dragan Simic (3):
phy: phy-rockchip-inno-usb2: Perform trivial code cleanups
phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better
phy: phy-rockchip-inno-usb2: Improve error handling while probing

 drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 63 ++++++++-----------
 1 file changed, 25 insertions(+), 38 deletions(-)
#1 lore raw

PATCH v2 1/3 phy: phy-rockchip-inno-usb2: Perform trivial code cleanups

Perform a few trivial code cleanups, e.g. to obey the reverse Christmas tree rule, to avoid use of "{ ... }" code blocks where they aren't really needed, or to avoid line wrapping by using the 100-column width better.

No intended functional changes are introduced by these code cleanups.

---
 drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 27 +++++++------------
 1 file changed, 9 insertions(+), 18 deletions(-)
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +9 -18
index 4f71373ae6e1..113bfc717ff0 100644
@@ -418,9 +418,9 @@ rockchip_usb2phy_clk480m_register(struct rockchip_usb2phy *rphy)
 
 static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy)
 {
-	int ret;
 	struct device_node *node = rphy->dev->of_node;
 	struct extcon_dev *edev;
+	int ret;
 
 	if (of_property_read_bool(node, "extcon")) {
 		edev = extcon_get_edev_by_phandle(rphy->dev, 0);
@@ -445,7 +445,6 @@ static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy)
 	}
 
 	rphy->edev = edev;
-
 	return 0;
 }
 
@@ -1327,21 +1326,19 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 	struct rockchip_usb2phy *rphy;
 	const struct rockchip_usb2phy_cfg *phy_cfgs;
 	unsigned int reg;
-	int index, ret;
+	int index = 0, ret;
 
 	rphy = devm_kzalloc(dev, sizeof(*rphy), GFP_KERNEL);
 	if (!rphy)
 		return -ENOMEM;
 
 	if (!dev->parent || !dev->parent->of_node) {
 		rphy->grf = syscon_regmap_lookup_by_phandle(np, "rockchip,usbgrf");
 		if (IS_ERR(rphy->grf)) {
 			dev_err(dev, "failed to locate usbgrf\n");
 			return PTR_ERR(rphy->grf);
 		}
-	}
-
-	else {
+	} else {
 		rphy->grf = syscon_node_to_regmap(dev->parent->of_node);
 		if (IS_ERR(rphy->grf))
 			return PTR_ERR(rphy->grf);
@@ -1358,16 +1355,14 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 	}
 
 	if (of_property_read_u32_index(np, "reg", 0, &reg)) {
-		dev_err(dev, "the reg property is not assigned in %pOFn node\n",
-			np);
+		dev_err(dev, "the reg property is not assigned in %pOFn node\n", np);
 		return -EINVAL;
 	}
 
 	/* support address_cells=2 */
 	if (of_property_count_u32_elems(np, "reg") > 2 && reg == 0) {
 		if (of_property_read_u32_index(np, "reg", 1, &reg)) {
-			dev_err(dev, "the reg property is not assigned in %pOFn node\n",
-				np);
+			dev_err(dev, "the reg property is not assigned in %pOFn node\n", np);
 			return -EINVAL;
 		}
 	}
@@ -1386,8 +1381,7 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 	if (ret)
 		return ret;
 
-	/* find out a proper config which can be matched with dt. */
-	index = 0;
+	/* find a proper config that can be matched with the DT */
 	do {
 		if (phy_cfgs[index].reg == reg) {
 			rphy->phy_cfg = &phy_cfgs[index];
@@ -1407,10 +1401,9 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 		return PTR_ERR(rphy->phy_reset);
 
 	rphy->clk = devm_clk_get_optional_enabled(dev, "phyclk");
-	if (IS_ERR(rphy->clk)) {
+	if (IS_ERR(rphy->clk))
 		return dev_err_probe(&pdev->dev, PTR_ERR(rphy->clk),
 				     "failed to get phyclk\n");
-	}
 
 	ret = rockchip_usb2phy_clk480m_register(rphy);
 	if (ret) {
@@ -1446,13 +1439,11 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 
 		/* initialize otg/host port separately */
 		if (of_node_name_eq(child_np, "host-port")) {
-			ret = rockchip_usb2phy_host_port_init(rphy, rport,
-							      child_np);
+			ret = rockchip_usb2phy_host_port_init(rphy, rport, child_np);
 			if (ret)
 				goto put_child;
 		} else {
-			ret = rockchip_usb2phy_otg_port_init(rphy, rport,
-							     child_np);
+			ret = rockchip_usb2phy_otg_port_init(rphy, rport, child_np);
 			if (ret)
 				goto put_child;
 		}
maintainer Reviewed-by on v2 1/3
#2 lore raw

Re: [PATCH v2 1/3] phy: phy-rockchip-inno-usb2: Perform trivial code cleanups

Am Mittwoch, 21. August 2024, 09:37:53 CEST schrieb Dragan Simic:

Quoted 9 lines

Perform a few trivial code cleanups, e.g. to obey the reverse Christmas tree rule, to avoid use of "{ ... }" code blocks where they aren't really needed, or to avoid line wrapping by using the 100-column width better.

No intended functional changes are introduced by these code cleanups.

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy) · line 449

Quoted 6 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -1
@@ -445,7 +445,6 @@ static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy)
 	}
 
 	rphy->edev = edev;
-
 	return 0;
 }

depending on how nitpicky we want to be, I'm not very fond of this change. Assigning the extcon-dev and returning "0" are two different actions, and I think most drivers do actually use a blank line between those.

But that really is just a style preference for me, so anyway

#3 lore raw

Re: [PATCH v2 1/3] phy: phy-rockchip-inno-usb2: Perform trivial code cleanups

Hello Heiko,

Quoted 22 lines · Heiko Stübner

Am Mittwoch, 21. August 2024, 09:37:53 CEST schrieb Dragan Simic:

Quoted 9 lines

Perform a few trivial code cleanups, e.g. to obey the reverse Christmas tree rule, to avoid use of "{ ... }" code blocks where they aren't really needed, or to avoid line wrapping by using the 100-column width better.

No intended functional changes are introduced by these code cleanups.

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_extcon_register(struct · line 450

Quoted 7 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -1
@@ -445,7 +445,6 @@ static int rockchip_usb2phy_extcon_register(struct 
 rockchip_usb2phy *rphy)
 	}
 
 	rphy->edev = edev;
-
 	return 0;
 }

depending on how nitpicky we want to be, I'm not very fond of this change. Assigning the extcon-dev and returning "0" are two different actions, and I think most drivers do actually use a blank line between those.

Quoted 3 lines

But that really is just a style preference for me, so anyway

Thanks for your review; I had similar thoughts, believe it or not. :) If there will be v3, I'll bring the empty separator line back.

#4 lore raw

PATCH v2 2/3 phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better

In reply to Dragan Simic

Return the actual error code upon failure to allocate extcon device, instead of hardcoding -ENOMEM. Use dev_err_probe() to also log appropriate messages, which is fine because the containing function is used in the probe path.

---
 drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +2 -1
index 113bfc717ff0..05af46dda11d 100644
@@ -435,7 +435,8 @@ static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy)
 						rockchip_usb2phy_extcon_cable);
 
 		if (IS_ERR(edev))
-			return -ENOMEM;
+			return dev_err_probe(rphy->dev, PTR_ERR(edev),
+					     "failed to allocate extcon device\n");
 
 		ret = devm_extcon_dev_register(rphy->dev, edev);
 		if (ret) {
maintainer Reviewed-by on v2 2/3
#5 lore raw

Re: [PATCH v2 2/3] phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better

Am Mittwoch, 21. August 2024, 09:37:54 CEST schrieb Dragan Simic:

Quoted 5 lines

Return the actual error code upon failure to allocate extcon device, instead of hardcoding -ENOMEM. Use dev_err_probe() to also log appropriate messages, which is fine because the containing function is used in the probe path.

How did I help with this? :-D

#6 lore raw

Re: [PATCH v2 2/3] phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better

Quoted 10 lines · Heiko Stübner

Am Mittwoch, 21. August 2024, 09:37:54 CEST schrieb Dragan Simic:

Quoted 5 lines

Return the actual error code upon failure to allocate extcon device, instead of hardcoding -ENOMEM. Use dev_err_probe() to also log appropriate messages, which is fine because the containing function is used in the probe path.

How did I help with this? :-D

Thanks for your review. Well, you suggested that dev_err_probe() is used. :)

#7 lore raw

Re: [PATCH v2 2/3] phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better

In reply to Dragan Simic

Le 21/08/2024 à 09:37, Dragan Simic a écrit :

Quoted 15 lines

Return the actual error code upon failure to allocate extcon device, instead of hardcoding -ENOMEM. Use dev_err_probe() to also log appropriate messages, which is fine because the containing function is used in the probe path.

---
  drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
index 113bfc717ff0..05af46dda11d 100644

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy) · line 439

Quoted 6 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +2 -1
@@ -435,7 +435,8 @@ static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy)
  						rockchip_usb2phy_extcon_cable);
 
  		if (IS_ERR(edev))
-			return -ENOMEM;
+			return dev_err_probe(rphy->dev, PTR_ERR(edev),
+					     "failed to allocate extcon device\n");

Returning PTR_ERR(edev) may make sense, but I don't think that adding a dev_err_probe() really helps.

devm_extcon_dev_allocate() can only return -EINVAL if it 2nd argument is NULL. It is trivial to see that it can't happen here, rockchip_usb2phy_extcon_cable is a global variable.

in all other cases, it returns -ENOMEM because of a failed memory allocation. In this case, usually it is not needed to log anything because it is already loud enough.

Just my 2c.

CJ

Quoted 5 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
  		ret = devm_extcon_dev_register(rphy->dev, edev);
  		if (ret) {
 
Signature · 3 lines
linux-phy mailing list
[email protected]
https://lists.infradead.org/mailman/listinfo/linux-phy
#8 lore raw

Re: [PATCH v2 2/3] phy: phy-rockchip-inno-usb2: Handle failed extcon allocation better

Hello Christophe,

Quoted 26 lines · Christophe JAILLET

Le 21/08/2024 à 09:37, Dragan Simic a écrit :

Quoted 15 lines

Return the actual error code upon failure to allocate extcon device, instead of hardcoding -ENOMEM. Use dev_err_probe() to also log appropriate messages, which is fine because the containing function is used in the probe path.

---
  drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
index 113bfc717ff0..05af46dda11d 100644

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_extcon_register(struct · line 439

Quoted 6 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +2 -1
@@ -435,7 +435,8 @@ static int rockchip_usb2phy_extcon_register(struct 
 rockchip_usb2phy *rphy)
  						rockchip_usb2phy_extcon_cable);
    		if (IS_ERR(edev))
-			return -ENOMEM;
+			return dev_err_probe(rphy->dev, PTR_ERR(edev),
+					     "failed to allocate extcon device\n");

Returning PTR_ERR(edev) may make sense, but I don't think that adding a dev_err_probe() really helps.

Quoted 7 lines

devm_extcon_dev_allocate() can only return -EINVAL if it 2nd argument is NULL. It is trivial to see that it can't happen here, rockchip_usb2phy_extcon_cable is a global variable.

in all other cases, it returns -ENOMEM because of a failed memory allocation. In this case, usually it is not needed to log anything because it is already loud enough.

True, using dev_err_probe() in this case is somewhat redundant, but it falls under the "still fine to use" category, [1] because the error code passed to dev_err_probe() is received from another function that was invoked.

On the other hand, another patch in this series tries to be strict in another direction, by not using dev_err_probe() where the error code passed to it is basically hardcoded. [2]

I hope this makes sense.

[1] https://lore.kernel.org/linux-phy/[email protected]/T/#mc3af7d24e31ed885732e6f26ca0d107b157d478b
[2] https://lore.kernel.org/linux-phy/[email protected]/T/#ma26b614412787814dab7923987b8c814f7e86beb

Quoted 4 lines
Quoted 4 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
    		ret = devm_extcon_dev_register(rphy->dev, edev);
  		if (ret) {
 
#9 lore raw

PATCH v2 3/3 phy: phy-rockchip-inno-usb2: Improve error handling while probing

In reply to Dragan Simic

Improve error handling in the probe path by using function dev_err_probe() where appropriate, and by no longer using it rather pointlessly in one place that actually produces a single, hardcoded error code.

---
 drivers/phy/rockchip/phy-rockchip-inno-usb2.c | 33 ++++++++-----------
 1 file changed, 14 insertions(+), 19 deletions(-)
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +14 -19
index 05af46dda11d..30b6134365ec 100644
@@ -424,25 +424,22 @@ static int rockchip_usb2phy_extcon_register(struct rockchip_usb2phy *rphy)
 
 	if (of_property_read_bool(node, "extcon")) {
 		edev = extcon_get_edev_by_phandle(rphy->dev, 0);
-		if (IS_ERR(edev)) {
-			if (PTR_ERR(edev) != -EPROBE_DEFER)
-				dev_err(rphy->dev, "Invalid or missing extcon\n");
-			return PTR_ERR(edev);
-		}
+		if (IS_ERR(edev))
+			return dev_err_probe(rphy->dev, PTR_ERR(edev),
+					     "invalid or missing extcon\n");
 	} else {
 		/* Initialize extcon device */
 		edev = devm_extcon_dev_allocate(rphy->dev,
 						rockchip_usb2phy_extcon_cable);
 
 		if (IS_ERR(edev))
 			return dev_err_probe(rphy->dev, PTR_ERR(edev),
 					     "failed to allocate extcon device\n");
 
 		ret = devm_extcon_dev_register(rphy->dev, edev);
-		if (ret) {
-			dev_err(rphy->dev, "failed to register extcon device\n");
-			return ret;
-		}
+		if (ret)
+			return dev_err_probe(rphy->dev, ret,
+					     "failed to register extcon device\n");
 	}
 
 	rphy->edev = edev;
@@ -1375,8 +1372,10 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 	rphy->irq = platform_get_irq_optional(pdev, 0);
 	platform_set_drvdata(pdev, rphy);
 
-	if (!phy_cfgs)
-		return dev_err_probe(dev, -EINVAL, "phy configs are not assigned!\n");
+	if (!phy_cfgs) {
+		dev_err(dev, "phy configs are not assigned\n");
+		return -EINVAL;
+	}
 
 	ret = rockchip_usb2phy_extcon_register(rphy);
 	if (ret)
@@ -1407,10 +1406,8 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 				     "failed to get phyclk\n");
 
 	ret = rockchip_usb2phy_clk480m_register(rphy);
-	if (ret) {
-		dev_err(dev, "failed to register 480m output clock\n");
-		return ret;
-	}
+	if (ret)
+		return dev_err_probe(dev, ret, "failed to register 480m output clock\n");
 
 	if (rphy->phy_cfg->phy_tuning) {
 		ret = rphy->phy_cfg->phy_tuning(rphy);
@@ -1430,8 +1427,7 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 
 		phy = devm_phy_create(dev, child_np, &rockchip_usb2phy_ops);
 		if (IS_ERR(phy)) {
-			dev_err_probe(dev, PTR_ERR(phy), "failed to create phy\n");
-			ret = PTR_ERR(phy);
+			ret = dev_err_probe(dev, PTR_ERR(phy), "failed to create phy\n");
 			goto put_child;
 		}
 
@@ -1466,8 +1462,7 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 						"rockchip_usb2phy",
 						rphy);
 		if (ret) {
-			dev_err(rphy->dev,
-				"failed to request usb2phy irq handle\n");
+			dev_err_probe(rphy->dev, ret, "failed to request usb2phy irq handle\n");
 			goto put_child;
 		}
 	}
maintainer
#10 lore raw

Re: [PATCH v2 3/3] phy: phy-rockchip-inno-usb2: Improve error handling while probing

Am Mittwoch, 21. August 2024, 09:37:55 CEST schrieb Dragan Simic:

Quoted 9 lines

Improve error handling in the probe path by using function dev_err_probe() where appropriate, and by no longer using it rather pointlessly in one place that actually produces a single, hardcoded error code.

drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
@@ -1375,8 +1372,10 @@ static int rockchip_usb2phy_probe(struct platform_device *pdev)
 	rphy->irq = platform_get_irq_optional(pdev, 0);
 	platform_set_drvdata(pdev, rphy);

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_probe(struct platform_device *pdev) · line 1381

Quoted 10 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +4 -2
 
-	if (!phy_cfgs)
-		return dev_err_probe(dev, -EINVAL, "phy configs are not assigned!\n");
+	if (!phy_cfgs) {
+		dev_err(dev, "phy configs are not assigned\n");
+		return -EINVAL;
+	}
 
 	ret = rockchip_usb2phy_extcon_register(rphy);
 	if (ret)

I really don't understand the rationale here. Using dev_err_probe here is just fine and with that change you just introduce more lines of code for exactly the same functionality?

#11 lore raw

Re: [PATCH v2 3/3] phy: phy-rockchip-inno-usb2: Improve error handling while probing

Quoted 20 lines · Heiko Stübner

Am Mittwoch, 21. August 2024, 09:37:55 CEST schrieb Dragan Simic:

Quoted 19 lines

Improve error handling in the probe path by using function dev_err_probe() where appropriate, and by no longer using it rather pointlessly in one place that actually produces a single, hardcoded error code.

drivers/phy/rockchip/phy-rockchip-inno-usb2.c +4 -2
@@ -1375,8 +1372,10 @@ static int rockchip_usb2phy_probe(struct 
 platform_device *pdev)
 	rphy->irq = platform_get_irq_optional(pdev, 0);
 	platform_set_drvdata(pdev, rphy);
 
-	if (!phy_cfgs)
-		return dev_err_probe(dev, -EINVAL, "phy configs are not 
 assigned!\n");
+	if (!phy_cfgs) {
+		dev_err(dev, "phy configs are not assigned\n");
+		return -EINVAL;
+	}
 
Quoted 6 lines

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_probe(struct platform_device *pdev) · line 1381

Quoted 2 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
 	ret = rockchip_usb2phy_extcon_register(rphy);
 	if (ret)

I really don't understand the rationale here. Using dev_err_probe here is just fine and with that change you just introduce more lines of code for exactly the same functionality?

As we know, dev_err_probe() decides how to log the received error message based on the error code it receives, but in this case the error code is hardcoded as -EINVAL. Thus, in this case it isn't about keeping the LoC count a bit lower, but about using dev_err() where the resulting outcome of error logging is aleady known, and where logging the error code actually isn't helpful, because it's hardcoded and the logged error message already tells everything about the error condition.

In other words, it's about being as precise as possible when deciding between dev_err() and dev_err_probe(), in both directions. I hope it makes sense.

maintainer
#12 lore raw

Re: [PATCH v2 3/3] phy: phy-rockchip-inno-usb2: Improve error handling while probing

Am Mittwoch, 21. August 2024, 11:09:03 CEST schrieb Dragan Simic:

Quoted 41 lines · Heiko Stübner
Quoted 22 lines

Am Mittwoch, 21. August 2024, 09:37:55 CEST schrieb Dragan Simic:

Quoted 21 lines

Improve error handling in the probe path by using function dev_err_probe()
where appropriate, and by no longer using it rather pointlessly in one place
that actually produces a single, hardcoded error code.

drivers/phy/rockchip/phy-rockchip-inno-usb2.c +4 -2
@@ -1375,8 +1372,10 @@ static int rockchip_usb2phy_probe(struct 
 platform_device *pdev)
 	rphy->irq = platform_get_irq_optional(pdev, 0);
 	platform_set_drvdata(pdev, rphy);
 
-	if (!phy_cfgs)
-		return dev_err_probe(dev, -EINVAL, "phy configs are not 
 assigned!\n");
+	if (!phy_cfgs) {
+		dev_err(dev, "phy configs are not assigned\n");
+		return -EINVAL;
+	}
 
Quoted 6 lines

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_probe(struct platform_device *pdev) · line 1381

Quoted 2 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
 	ret = rockchip_usb2phy_extcon_register(rphy);
 	if (ret)

I really don't understand the rationale here. Using dev_err_probe here is just fine and with that change you just introduce more lines of code for exactly the same functionality?

As we know, dev_err_probe() decides how to log the received error message
based on the error code it receives, but in this case the error code is hardcoded as -EINVAL. Thus, in this case it isn't about keeping the LoC count a bit lower, but about using dev_err() where the resulting outcome of error logging is aleady known, and where logging the error code actually
isn't helpful, because it's hardcoded and the logged error message already
tells everything about the error condition.

Quoted 4 lines

In other words, it's about being as precise as possible when deciding between
dev_err() and dev_err_probe(), in both directions. I hope it makes sense.

I'd disagree a bit, using one format only creates a way nicer pattern in the driver, by not mixing different styles.

dev_err_probe documentation seems to agree [0], by stating:

"Using this helper in your probe function is totally fine even if @err is
known to never be -EPROBE_DEFER.
The benefit compared to a normal dev_err() is the standardized format of the error code, it being emitted symbolically (i.e. you get "EAGAIN" instead of "-35") and the fact that the error code is returned which allows more compact error paths."

[0] https://elixir.bootlin.com/linux/v6.10.6/source/drivers/base/core.c#L5009

Signature · 3 lines
linux-phy mailing list
[email protected]
https://lists.infradead.org/mailman/listinfo/linux-phy
#13 lore raw

Re: [PATCH v2 3/3] phy: phy-rockchip-inno-usb2: Improve error handling while probing

Quoted 52 lines · Heiko Stübner

Am Mittwoch, 21. August 2024, 11:09:03 CEST schrieb Dragan Simic:

Quoted 41 lines · Heiko Stübner
Quoted 22 lines

Am Mittwoch, 21. August 2024, 09:37:55 CEST schrieb Dragan Simic:

Quoted 21 lines

Improve error handling in the probe path by using function dev_err_probe()
where appropriate, and by no longer using it rather pointlessly in one place
that actually produces a single, hardcoded error code.

drivers/phy/rockchip/phy-rockchip-inno-usb2.c +4 -2
@@ -1375,8 +1372,10 @@ static int rockchip_usb2phy_probe(struct
 platform_device *pdev)
 	rphy->irq = platform_get_irq_optional(pdev, 0);
 	platform_set_drvdata(pdev, rphy);
 
-	if (!phy_cfgs)
-		return dev_err_probe(dev, -EINVAL, "phy configs are not
 assigned!\n");
+	if (!phy_cfgs) {
+		dev_err(dev, "phy configs are not assigned\n");
+		return -EINVAL;
+	}
 
Quoted 6 lines

drivers/phy/rockchip/phy-rockchip-inno-usb2.c static int rockchip_usb2phy_probe(struct platform_device *pdev) · line 1381

Quoted 2 lines
drivers/phy/rockchip/phy-rockchip-inno-usb2.c +0 -0
 	ret = rockchip_usb2phy_extcon_register(rphy);
 	if (ret)

I really don't understand the rationale here. Using dev_err_probe here is just fine and with that change you just introduce more lines of code for exactly the same functionality?

As we know, dev_err_probe() decides how to log the received error message
based on the error code it receives, but in this case the error code is hardcoded as -EINVAL. Thus, in this case it isn't about keeping the LoC count a bit lower, but about using dev_err() where the resulting outcome of error logging is aleady known, and where logging the error code actually
isn't helpful, because it's hardcoded and the logged error message already
tells everything about the error condition.

Quoted 4 lines

In other words, it's about being as precise as possible when deciding between
dev_err() and dev_err_probe(), in both directions. I hope it makes sense.

I'd disagree a bit, using one format only creates a way nicer pattern in the driver, by not mixing different styles.

dev_err_probe documentation seems to agree [0], by stating:

Quoted 8 lines

"Using this helper in your probe function is totally fine even if @err is
known to never be -EPROBE_DEFER.
The benefit compared to a normal dev_err() is the standardized format of the error code, it being emitted symbolically (i.e. you get
"EAGAIN"
instead of "-35") and the fact that the error code is returned which
allows
more compact error paths."

Yes, I saw that already in the documentation. Though, it might be debatable does hardcoding the passed error code to some value qualifies as knowing that it can't be -EPROBE_DEFER. The way I read that part of the documentation is that using dev_err_probe() is fine without going into the implementation of the previously invoked function that may fail, and researching can it actually return -EPROBE_DEFER or not. Also, the invoked function may change at some point in future and start returning -EPROBE_DEFER, but a hardcoded error code that's produced locally can't become changed that way.

In addition to that, we already have at least a couple of instances [1][2] in the same function in which dev_err() is used when the error code is hardcoded, so there's actually already another pattern to follow.

I know that replacing dev_err_probe() with dev_err() may look strange in a patch that mostly performs the opposite replacement, but the patch just tries to be strict and precise, and to follow other examples of how dev_err() is already used in the same function when the error code is produced locally instead of being received from another invoked function.

Quoted 1 line

[0] https://elixir.bootlin.com/linux/v6.10.6/source/drivers/base/core.c#L5009

[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/rockchip/phy-rockchip-inno-usb2.c?h=v6.11-rc4#n1361
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/rockchip/phy-rockchip-inno-usb2.c?h=v6.11-rc4#n1369