Skip to contentSkip to conversation

rtc: msc313: Fix function prototype mismatch in msc313_rtc_probe()

PATCH Superseded Kees Cook 2 participants 4 messages Nov 18, 2022 – Dec 2, 2022

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Kees Cook

author
#0 lore raw

PATCH 1/1 rtc: msc313: Fix function prototype mismatch in msc313_rtc_probe()

With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG), indirect call targets are validated against the expected function pointer prototype to make sure the call target is valid to help mitigate ROP attacks. If they are not identical, there is a failure at run time, which manifests as either a kernel panic or thread getting killed.

msc313_rtc_probe() was passing clk_disable_unprepare() directly, which did not have matching prototypes for devm_add_action_or_reset()'s callback argument. Add a wrapper and remove the cast.

This was found as a result of Clang's new -Wcast-function-type-strict flag, which is more sensitive than the simpler -Wcast-function-type, which only checks for type width mismatches.

---
 drivers/rtc/rtc-msc313.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/rtc/rtc-msc313.c +8 -1
index f3fde013c4b8..36e3e77f303e 100644
@@ -177,6 +177,13 @@ static irqreturn_t msc313_rtc_interrupt(s32 irq, void *dev_id)
 	return IRQ_HANDLED;
 }
 
+static void msc313_clk_disable_unprepare(void *data)
+{
+	struct clk *clk = data;
+
+	clk_disable_unprepare(clk);
+}
+
 static int msc313_rtc_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -224,7 +231,7 @@ static int msc313_rtc_probe(struct platform_device *pdev)
 		return ret;
 	}
 
-	ret = devm_add_action_or_reset(dev, (void (*) (void *))clk_disable_unprepare, clk);
+	ret = devm_add_action_or_reset(dev, msc313_clk_disable_unprepare, clk);
 	if (ret)
 		return ret;
 
Signature · 1 line
2.34.1
#1 lore raw

Le 19/11/2022 à 00:31, Kees Cook a écrit :

Quoted 47 lines

With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG), indirect call targets are validated against the expected function pointer prototype to make sure the call target is valid to help mitigate ROP attacks. If they are not identical, there is a failure at run time, which manifests as either a kernel panic or thread getting killed.

msc313_rtc_probe() was passing clk_disable_unprepare() directly, which did not have matching prototypes for devm_add_action_or_reset()'s callback argument. Add a wrapper and remove the cast.

This was found as a result of Clang's new -Wcast-function-type-strict flag, which is more sensitive than the simpler -Wcast-function-type, which only checks for type width mismatches.

---
  drivers/rtc/rtc-msc313.c | 9 ++++++++-
  1 file changed, 8 insertions(+), 1 deletion(-)
drivers/rtc/rtc-msc313.c +7 -0
index f3fde013c4b8..36e3e77f303e 100644
@@ -177,6 +177,13 @@ static irqreturn_t msc313_rtc_interrupt(s32 irq, void *dev_id)
  	return IRQ_HANDLED;
  }
 
+static void msc313_clk_disable_unprepare(void *data)
+{
+	struct clk *clk = data;
+
+	clk_disable_unprepare(clk);
+}
+
  static int msc313_rtc_probe(struct platform_device *pdev)
  {
  	struct device *dev = &pdev->dev;

drivers/rtc/rtc-msc313.c static int msc313_rtc_probe(struct platform_device *pdev) · line 237

Quoted 8 lines
drivers/rtc/rtc-msc313.c +1 -1
@@ -224,7 +231,7 @@ static int msc313_rtc_probe(struct platform_device *pdev)
  		return ret;
  	}
 
-	ret = devm_add_action_or_reset(dev, (void (*) (void *))clk_disable_unprepare, clk);
+	ret = devm_add_action_or_reset(dev, msc313_clk_disable_unprepare, clk);
  	if (ret)
  		return ret;
 

Hi,

another way to fix it, is to use devm_clk_get_enabled().

It removes some LoC instead of introducing some new ones and saves a few bytes of memory.

CJ

author
#2 lore raw
Quoted 56 lines · Christophe JAILLET

Le 19/11/2022 à 00:31, Kees Cook a écrit :

Quoted 46 lines

With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG), indirect call targets are validated against the expected function pointer prototype to make sure the call target is valid to help mitigate ROP attacks. If they are not identical, there is a failure at run time, which manifests as either a kernel panic or thread getting killed.

msc313_rtc_probe() was passing clk_disable_unprepare() directly, which did not have matching prototypes for devm_add_action_or_reset()'s callback argument. Add a wrapper and remove the cast.

This was found as a result of Clang's new -Wcast-function-type-strict flag, which is more sensitive than the simpler -Wcast-function-type, which only checks for type width mismatches.

---
  drivers/rtc/rtc-msc313.c | 9 ++++++++-
  1 file changed, 8 insertions(+), 1 deletion(-)
drivers/rtc/rtc-msc313.c +7 -0
index f3fde013c4b8..36e3e77f303e 100644
@@ -177,6 +177,13 @@ static irqreturn_t msc313_rtc_interrupt(s32 irq, void *dev_id)
  	return IRQ_HANDLED;
  }
+static void msc313_clk_disable_unprepare(void *data)
+{
+	struct clk *clk = data;
+
+	clk_disable_unprepare(clk);
+}
+
  static int msc313_rtc_probe(struct platform_device *pdev)
  {
  	struct device *dev = &pdev->dev;

drivers/rtc/rtc-msc313.c static int msc313_rtc_probe(struct platform_device *pdev) · line 235

Quoted 6 lines
drivers/rtc/rtc-msc313.c +1 -1
@@ -224,7 +231,7 @@ static int msc313_rtc_probe(struct platform_device *pdev)
  		return ret;
  	}
-	ret = devm_add_action_or_reset(dev, (void (*) (void *))clk_disable_unprepare, clk);
+	ret = devm_add_action_or_reset(dev, msc313_clk_disable_unprepare, clk);
  	if (ret)
  		return ret;

Hi,

Quoted 4 lines

another way to fix it, is to use devm_clk_get_enabled().

It removes some LoC instead of introducing some new ones and saves a few bytes of memory.

Hrm, I'm not familiar with the clk stuff here -- how do I use it? Should it just be like this? (The NULL argument is ok?)

drivers/rtc/rtc-msc313.c +1 -11
index f3fde013c4b8..8d7737e0e2e0 100644
@@ -212,22 +212,12 @@ static int msc313_rtc_probe(struct platform_device *pdev)
 		return ret;
 	}
 
-	clk = devm_clk_get(dev, NULL);
+	clk = devm_clk_get_enabled(dev, NULL);
 	if (IS_ERR(clk)) {
 		dev_err(dev, "No input reference clock\n");
 		return PTR_ERR(clk);
 	}
 
-	ret = clk_prepare_enable(clk);
-	if (ret) {
-		dev_err(dev, "Failed to enable the reference clock, %d\n", ret);
-		return ret;
-	}
-
-	ret = devm_add_action_or_reset(dev, (void (*) (void *))clk_disable_unprepare, clk);
-	if (ret)
-		return ret;
-
 	rate = clk_get_rate(clk);
 	writew(rate & 0xFFFF, priv->rtc_base + REG_RTC_FREQ_CW_L);
 	writew((rate >> 16) & 0xFFFF, priv->rtc_base + REG_RTC_FREQ_CW_H);
Signature · 1 line
Kees Cook
#3 lore raw

Le 02/12/2022 à 05:46, Kees Cook a écrit :

Quoted 57 lines · Christophe JAILLET
Quoted 56 lines

Le 19/11/2022 à 00:31, Kees Cook a écrit :

Quoted 46 lines

With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG), indirect call targets are validated against the expected function pointer prototype to make sure the call target is valid to help mitigate ROP attacks. If they are not identical, there is a failure at run time, which manifests as either a kernel panic or thread getting killed.

msc313_rtc_probe() was passing clk_disable_unprepare() directly, which did not have matching prototypes for devm_add_action_or_reset()'s callback argument. Add a wrapper and remove the cast.

This was found as a result of Clang's new -Wcast-function-type-strict flag, which is more sensitive than the simpler -Wcast-function-type, which only checks for type width mismatches.

---
   drivers/rtc/rtc-msc313.c | 9 ++++++++-
   1 file changed, 8 insertions(+), 1 deletion(-)
drivers/rtc/rtc-msc313.c +7 -0
index f3fde013c4b8..36e3e77f303e 100644
@@ -177,6 +177,13 @@ static irqreturn_t msc313_rtc_interrupt(s32 irq, void *dev_id)
   	return IRQ_HANDLED;
   }
+static void msc313_clk_disable_unprepare(void *data)
+{
+	struct clk *clk = data;
+
+	clk_disable_unprepare(clk);
+}
+
   static int msc313_rtc_probe(struct platform_device *pdev)
   {
   	struct device *dev = &pdev->dev;

drivers/rtc/rtc-msc313.c static int msc313_rtc_probe(struct platform_device *pdev) · line 235

Quoted 6 lines
drivers/rtc/rtc-msc313.c +1 -1
@@ -224,7 +231,7 @@ static int msc313_rtc_probe(struct platform_device *pdev)
   		return ret;
   	}
-	ret = devm_add_action_or_reset(dev, (void (*) (void *))clk_disable_unprepare, clk);
+	ret = devm_add_action_or_reset(dev, msc313_clk_disable_unprepare, clk);
   	if (ret)
   		return ret;

Hi,

Quoted 7 lines
Quoted 4 lines

another way to fix it, is to use devm_clk_get_enabled().

It removes some LoC instead of introducing some new ones and saves a few bytes of memory.

Hrm, I'm not familiar with the clk stuff here -- how do I use it? Should it just be like this? (The NULL argument is ok?)

Yes, the code below looks fine.
NULL argument is ok.

See [1]. Both function are very similar. devm_clk_get_enabled() just pass some additional init() and exit() functions.

CJ

[1]: https://elixir.bootlin.com/linux/v6.1-rc7/source/drivers/clk/clk-devres.c#L63

Quoted 30 lines
drivers/rtc/rtc-msc313.c +1 -11
index f3fde013c4b8..8d7737e0e2e0 100644
@@ -212,22 +212,12 @@ static int msc313_rtc_probe(struct platform_device *pdev)
  		return ret;
  	}
 
-	clk = devm_clk_get(dev, NULL);
+	clk = devm_clk_get_enabled(dev, NULL);
  	if (IS_ERR(clk)) {
  		dev_err(dev, "No input reference clock\n");
  		return PTR_ERR(clk);
  	}
 
-	ret = clk_prepare_enable(clk);
-	if (ret) {
-		dev_err(dev, "Failed to enable the reference clock, %d\n", ret);
-		return ret;
-	}
-
-	ret = devm_add_action_or_reset(dev, (void (*) (void *))clk_disable_unprepare, clk);
-	if (ret)
-		return ret;
-
  	rate = clk_get_rate(clk);
  	writew(rate & 0xFFFF, priv->rtc_base + REG_RTC_FREQ_CW_L);
  	writew((rate >> 16) & 0xFFFF, priv->rtc_base + REG_RTC_FREQ_CW_H);