Skip to contentSkip to conversation

rtc: stm32: fix clock leak on pinctrl probe failure

PATCH In review Cong Nguyen 3 participants 3 messages Sep 18 – Sep 30

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 rtc: stm32: fix clock leak on pinctrl probe failure

A failing devm_pinctrl_register_and_init() or pinctrl_enable() in stm32_rtc_probe() returns directly instead of going through err:, leaking rtc_ck/pclk -- and since -EPROBE_DEFER is routine here, each deferred retry leaks another.

Route both through err:, and move the pinctrl setup earlier (right after stm32_rtc_init(), before the RTC device/IRQ are registered) so disabling the clocks there doesn't race a still-live IRQ.

Fixes 16ad2bc09efb ("rtc: stm32: add pinctrl and pinmux interfaces")
Assisted-by LLM

---
Changes in v2:
- Sashiko found that routing the pinctrl failures through err: (as v1 did) disables the clocks while the still-registered RTC device and alarm IRQ are live. Moved the pinctrl setup earlier instead of just moving where the leak was plugged.

 drivers/rtc/rtc-stm32.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)
drivers/rtc/rtc-stm32.c +14 -10
index d4ebf3eb54aa..457f66663ba5 100644
@@ -1137,6 +1137,20 @@ static int stm32_rtc_probe(struct platform_device *pdev)
 	if (ret)
 		goto err;
 
+	stm32_rtc_clean_outs(rtc);
+
+	ret = devm_pinctrl_register_and_init(&pdev->dev, &stm32_rtc_pdesc, rtc, &pctl);
+	if (ret) {
+		dev_err_probe(&pdev->dev, ret, "pinctrl register failed");
+		goto err;
+	}
+
+	ret = pinctrl_enable(pctl);
+	if (ret) {
+		dev_err_probe(&pdev->dev, ret, "pinctrl enable failed");
+		goto err;
+	}
+
 	rtc->irq_alarm = platform_get_irq(pdev, 0);
 	if (rtc->irq_alarm <= 0) {
 		ret = rtc->irq_alarm;
@@ -1172,16 +1186,6 @@ static int stm32_rtc_probe(struct platform_device *pdev)
 		goto err;
 	}
 
-	stm32_rtc_clean_outs(rtc);
-
-	ret = devm_pinctrl_register_and_init(&pdev->dev, &stm32_rtc_pdesc, rtc, &pctl);
-	if (ret)
-		return dev_err_probe(&pdev->dev, ret, "pinctrl register failed");
-
-	ret = pinctrl_enable(pctl);
-	if (ret)
-		return dev_err_probe(&pdev->dev, ret, "pinctrl enable failed");
-
 	/*
 	 * If INITS flag is reset (calendar year field set to 0x00), calendar
 	 * must be initialized
Signature · 1 line
2.25.1
#1 lore raw
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider: - [High] rtc: stm32: NULL pointer dereference in stm32_rtc_pinmux_action_lsco() - [High] rtc: stm32…

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] rtc: stm32: NULL pointer dereference in stm32_rtc_pinmux_action_lsco()
- [High] rtc: stm32: clk_lsco and OF provider leak on probe error
--

commit c5ee543f4626bb3238dc601e2915f1d27f3674f2
Author: Cong Nguyen <[email protected]>

rtc: stm32: fix clock leak on pinctrl probe failure

A failing devm_pinctrl_register_and_init() or pinctrl_enable() in stm32_rtc_probe() returns directly instead of going through err:, leaking rtc_ck/pclk -- and since -EPROBE_DEFER is routine here, each deferred retry leaks another.

Quoted 7 lines
drivers/rtc/rtc-stm32.c +0 -0
index d4ebf3eb54aa9..457f66663ba59 100644
@@ -1137,6 +1137,20 @@ static int stm32_rtc_probe(struct platform_device *pdev)
 	if (ret)
 		goto err;

drivers/rtc/rtc-stm32.c static int stm32_rtc_probe(struct platform_device *pdev) · line 1148

Quoted 10 lines
drivers/rtc/rtc-stm32.c +9 -0
 
+	stm32_rtc_clean_outs(rtc);
+
+	ret = devm_pinctrl_register_and_init(&pdev->dev, &stm32_rtc_pdesc, rtc, &pctl);
+	if (ret) {
+		dev_err_probe(&pdev->dev, ret, "pinctrl register failed");
+		goto err;
+	}
+
+	ret = pinctrl_enable(pctl);

[Severity: High]
Does this code trigger a null pointer dereference if the 'lsco' function is configured as a pinmux hog in the device tree?

Calling pinctrl_enable() synchronously claims hogs, which invokes the pinmux callback. If 'lsco' is claimed, stm32_rtc_pinmux_action_lsco() will be called:

stm32_rtc_pinmux_action_lsco() {
    struct device *dev = rtc->rtc_dev->dev.parent;

Since devm_rtc_device_register() hasn't been called yet in the new probe sequence, rtc->rtc_dev is still NULL, which will cause a kernel panic during probe.

Quoted 9 lines
drivers/rtc/rtc-stm32.c +5 -0
+	if (ret) {
+		dev_err_probe(&pdev->dev, ret, "pinctrl enable failed");
+		goto err;
+	}
+
 	rtc->irq_alarm = platform_get_irq(pdev, 0);
 	if (rtc->irq_alarm <= 0) {
 		ret = rtc->irq_alarm;

goto err;

[Severity: High]
Can a failure here leak the manually registered clock gate and OF provider?

If 'lsco' is claimed by pinctrl_enable() above, it registers a clock gate and an OF provider:

stm32_rtc_pinmux_action_lsco() {
    ...
    rtc->clk_lsco = clk_register_gate(dev, "rtc_lsco", ...);
    of_clk_add_provider(dev->of_node, of_clk_src_simple_get, rtc->clk_lsco);

Because pinctrl setup was moved earlier, any subsequent error such as platform_get_irq() failing will jump to the err: label. The err: cleanup path does not unregister rtc->clk_lsco or the OF provider. This will permanently leak these resources on routine errors like -EPROBE_DEFER, prevent successful re-probing, and can cause a use-after-free later since there is no of_clk_del_provider() call.

Quoted 1 line

}

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Reviewed-by on v2 1/1
#2 lore raw
In reply to Cong Nguyen
Quoted 9 lines · Cong Nguyen

A failing devm_pinctrl_register_and_init() or pinctrl_enable() in stm32_rtc_probe() returns directly instead of going through err:, leaking rtc_ck/pclk -- and since -EPROBE_DEFER is routine here, each deferred retry leaks another.

Route both through err:, and move the pinctrl setup earlier (right after stm32_rtc_init(), before the RTC device/IRQ are registered) so disabling the clocks there doesn't race a still-live IRQ.

Quoted 5 lines
Fixes 16ad2bc09efb ("rtc: stm32: add pinctrl and pinmux interfaces")
Assisted-by LLM

Yours,
Linus Walleij