Skip to contentSkip to conversation

hwrng: omap: Fix probe error path cleanup

PATCH In review Myeonghun Pak 2 participants 2 messages Jun 23 – Jul 5

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hwrng: omap: Fix probe error path cleanup

omap_rng_probe() enables runtime PM before acquiring and enabling the functional clocks. Several later error paths returned or unwound without undoing all state acquired so far.

If pm_runtime_resume_and_get() failed, the driver returned through the generic ioremap error label and left runtime PM enabled. If either clock lookup returned -EPROBE_DEFER, the function returned directly and skipped the runtime PM cleanup; the register clock defer path could also leave the already enabled functional clock prepared.

Route these failures through the existing unwind labels so each path only undoes resources that were acquired successfully. Keep the resume failure path limited to pm_runtime_disable(), and use the later labels only after the runtime PM usage count or clocks have been acquired.

This issue was identified during our ongoing static-analysis research while reviewing kernel code.

Fixes 61dc0a446e5d ("hwrng: omap - Fix assumption that runtime_get_sync will always succeed")
Fixes 43ec540e6f9b ("hwrng: omap - move clock related code to omap_rng_probe()")
Fixes b166be004491 ("hwrng: omap - Fix clock resource by adding a register clock")
---
 drivers/char/hw_random/omap-rng.c | 30 ++++++++++++++++++++----------
 1 file changed, 20 insertions(+), 10 deletions(-)
drivers/char/hw_random/omap-rng.c +20 -10
index 5e8b50f15d..a8c0b3dfb1 100644
@@ -455,32 +455,40 @@ static int omap_rng_probe(struct platform_device *pdev)
 	ret = pm_runtime_resume_and_get(&pdev->dev);
 	if (ret < 0) {
 		dev_err(&pdev->dev, "Failed to runtime_get device: %d\n", ret);
-		goto err_ioremap;
+		goto err_pm_disable;
 	}
 
 	priv->clk = devm_clk_get(&pdev->dev, NULL);
-	if (PTR_ERR(priv->clk) == -EPROBE_DEFER)
-		return -EPROBE_DEFER;
+	if (PTR_ERR(priv->clk) == -EPROBE_DEFER) {
+		ret = -EPROBE_DEFER;
+		goto err_pm_put;
+	}
 	if (!IS_ERR(priv->clk)) {
 		ret = clk_prepare_enable(priv->clk);
 		if (ret) {
 			dev_err(&pdev->dev,
 				"Unable to enable the clk: %d\n", ret);
-			goto err_register;
+			goto err_pm_put;
 		}
+	} else {
+		priv->clk = NULL;
 	}
 
 	priv->clk_reg = devm_clk_get(&pdev->dev, "reg");
-	if (PTR_ERR(priv->clk_reg) == -EPROBE_DEFER)
-		return -EPROBE_DEFER;
+	if (PTR_ERR(priv->clk_reg) == -EPROBE_DEFER) {
+		ret = -EPROBE_DEFER;
+		goto err_clk;
+	}
 	if (!IS_ERR(priv->clk_reg)) {
 		ret = clk_prepare_enable(priv->clk_reg);
 		if (ret) {
 			dev_err(&pdev->dev,
 				"Unable to enable the register clk: %d\n",
 				ret);
-			goto err_register;
+			goto err_clk;
 		}
+	} else {
+		priv->clk_reg = NULL;
 	}
 
 	ret = (dev->of_node) ? of_get_omap_rng_device_details(priv, pdev) :
@@ -498,12 +506,14 @@ static int omap_rng_probe(struct platform_device *pdev)
 	return 0;
 
 err_register:
+	clk_disable_unprepare(priv->clk_reg);
+err_clk:
+	clk_disable_unprepare(priv->clk);
+err_pm_put:
 	priv->base = NULL;
 	pm_runtime_put_sync(&pdev->dev);
+err_pm_disable:
 	pm_runtime_disable(&pdev->dev);
-
-	clk_disable_unprepare(priv->clk_reg);
-	clk_disable_unprepare(priv->clk);
 err_ioremap:
 	dev_err(dev, "initialization failed.\n");
 	return ret;
Signature · 1 line
2.47.1
maintainer
#1 lore raw
Quoted 25 lines · Myeonghun Pak

omap_rng_probe() enables runtime PM before acquiring and enabling the functional clocks. Several later error paths returned or unwound without undoing all state acquired so far.

If pm_runtime_resume_and_get() failed, the driver returned through the generic ioremap error label and left runtime PM enabled. If either clock lookup returned -EPROBE_DEFER, the function returned directly and skipped the runtime PM cleanup; the register clock defer path could also leave the already enabled functional clock prepared.

Route these failures through the existing unwind labels so each path only undoes resources that were acquired successfully. Keep the resume failure path limited to pm_runtime_disable(), and use the later labels only after the runtime PM usage count or clocks have been acquired.

This issue was identified during our ongoing static-analysis research while reviewing kernel code.

Fixes 61dc0a446e5d ("hwrng: omap - Fix assumption that runtime_get_sync will always succeed")
Fixes 43ec540e6f9b ("hwrng: omap - move clock related code to omap_rng_probe()")
Fixes b166be004491 ("hwrng: omap - Fix clock resource by adding a register clock")
Quoted 3 lines
---
 drivers/char/hw_random/omap-rng.c | 30 ++++++++++++++++++++----------
 1 file changed, 20 insertions(+), 10 deletions(-)

Patch applied. Thanks.

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt