Skip to contentSkip to conversation

hwmon: (lm63) Convert macro to function to avoid TOCTOU

PATCH In review Gui-Dong Han 2 participants 6 messages Apr 16 Hardware monitoring

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 hwmon: (lm63) Convert macro to function to avoid TOCTOU

The macro FAN_FROM_REG evaluates its argument multiple times. When used in lockless code accessing shared driver data, this can cause a Time-of-Check to Time-of-Use (TOCTOU) race and potentially a divide-by-zero error.

Convert the macro to a static function so that the register value is evaluated only once.

Check the other conversion macros in the driver as well. Keep them unchanged because they either do not evaluate arguments multiple times or are only used from locked code paths.

---
While learning the hwmon driver code, I found a few more potential TOCTOU problems in drivers still using the older non-_with_info() APIs. Fix them.

---
 drivers/hwmon/lm63.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
drivers/hwmon/lm63.c +9 -3
index 035176a98ce9..da132b267c58 100644
@@ -109,8 +109,14 @@ static const unsigned short normal_i2c[] = { 0x18, 0x4c, 0x4e, I2C_CLIENT_END };
  * adapted accordingly.
  */
 
-#define FAN_FROM_REG(reg)	((reg) == 0xFFFC || (reg) == 0 ? 0 : \
-				 5400000 / (reg))
+static int fan_from_reg(int reg)
+{
+	if (reg == 0xFFFC || reg == 0)
+		return 0;
+
+	return 5400000 / reg;
+}
+
 #define FAN_TO_REG(val)		((val) <= 82 ? 0xFFFC : \
 				 (5400000 / (val)) & 0xFFFC)
 #define TEMP8_FROM_REG(reg)	((reg) * 1000)
@@ -333,7 +339,7 @@ static ssize_t show_fan(struct device *dev, struct device_attribute *devattr,
 {
 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 	struct lm63_data *data = lm63_update_device(dev);
-	return sprintf(buf, "%d\n", FAN_FROM_REG(data->fan[attr->index]));
+	return sprintf(buf, "%d\n", fan_from_reg(data->fan[attr->index]));
 }
 
 static ssize_t set_fan(struct device *dev, struct device_attribute *dummy,
Signature · 1 line
2.43.0
#1 lore raw

PATCH 2/2 hwmon: (lm63) Add locking to avoid TOCTOU

The functions show_pwm1() and show_temp11() access shared cached data multiple times without holding the update lock. This can cause TOCTOU races if the cached values change between the checks and the later calculations.

Those cached values are updated in lm63_update_device(). For example, show_pwm1() may read data->pwm1[nr] and data->pwm1_freq while a concurrent lm63_update_device() refresh is updating them, resulting in a mixed old/new snapshot.

Hold the update lock across the whole read and calculation sequence so that the values remain stable.

Check the other functions in the driver as well. Keep them unchanged because they either do not access shared cached values multiple times or already do so under lock.

Fixes 1da177e4c3f4 ("Linux-2.6.12-rc2")
Fixes e872c91e726e ("hwmon: (lm63) Add support for unsigned upper temperature limits")

---
While learning the hwmon driver code, I found a few more potential TOCTOU problems in drivers still using the older non-_with_info() APIs. Fix them.

---
 drivers/hwmon/lm63.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/hwmon/lm63.c +7 -1
index da132b267c58..25b871695cb3 100644
@@ -372,12 +372,14 @@ static ssize_t show_pwm1(struct device *dev, struct device_attribute *devattr,
 	int nr = attr->index;
 	int pwm;
 
+	mutex_lock(&data->update_lock);
 	if (data->pwm_highres)
 		pwm = data->pwm1[nr];
 	else
 		pwm = data->pwm1[nr] >= 2 * data->pwm1_freq ?
 		       255 : (data->pwm1[nr] * 255 + data->pwm1_freq) /
 		       (2 * data->pwm1_freq);
+	mutex_unlock(&data->update_lock);
 
 	return sprintf(buf, "%d\n", pwm);
 }
@@ -535,6 +537,7 @@ static ssize_t show_temp11(struct device *dev, struct device_attribute *devattr,
 	int nr = attr->index;
 	int temp;
 
+	mutex_lock(&data->update_lock);
 	if (!nr) {
 		/*
 		 * Use unsigned temperature unless its value is zero.
@@ -550,7 +553,10 @@ static ssize_t show_temp11(struct device *dev, struct device_attribute *devattr,
 		else
 			temp = TEMP11_FROM_REG(data->temp11[nr]);
 	}
-	return sprintf(buf, "%d\n", temp + data->temp2_offset);
+	temp += data->temp2_offset;
+	mutex_unlock(&data->update_lock);
+
+	return sprintf(buf, "%d\n", temp);
 }
 
 static ssize_t set_temp11(struct device *dev, struct device_attribute *devattr,
Signature · 1 line
2.43.0
#2 lore raw

Re: [PATCH 2/2] hwmon: (lm63) Add locking to avoid TOCTOU

Sashiko AI review found 3 potential issue(s): - [High] Pre-existing Torn Read Race Condition in show_fan() causes massive fan speed spikes - [High] Pre-existing Uninitialized Cache Read in…

Sashiko AI review found 3 potential issue(s):
- [High] Pre-existing Torn Read Race Condition in show_fan() causes massive fan speed spikes
- [High] Pre-existing Uninitialized Cache Read in temp2_crit_hyst_store() corrupts hardware thermal limits
- [Medium] Incomplete Fix: Unaddressed TOCTOU Races in temp2_crit_hyst_show() and show_lut_temp_hyst() --

commit 1233dd514546736dfbbf09561bf360360c5933ca
Author: Gui-Dong Han <[email protected]>

hwmon: (lm63) Add locking to avoid TOCTOU

This commit adds locking to show_pwm1() and show_temp11() to prevent TOCTOU races when reading multiple cached values. It also states that other functions in the driver were checked for similar issues.

Quoted 3 lines

Check the other functions in the driver as well. Keep them unchanged because they either do not access shared cached values multiple times or already do so under lock.

Is this description accurate?

Looking at temp2_crit_hyst_show(), it calculates its value by reading both data->temp8[2] and data->temp2_crit_hyst. Similarly, show_lut_temp_hyst() reads data->temp8[attr->index] and data->lut_temp_hyst. Because neither function holds data->update_lock, can a concurrent cache update cause them to fetch a mixed snapshot (e.g., a newly updated base temperature but a stale hysteresis offset), resulting in corrupted absolute hysteresis values returned to userspace?

Also, looking at show_fan(), it accesses data->fan[attr->index] without holding data->update_lock. In lm63_update_device(), data->fan[0] is populated via two separate non-atomic assignments. Can show_fan() read the intermediate torn state where the MSB is still 0, causing the calculation (5400000 / reg) to report massive, incorrect fan speed spikes?

Finally, in temp2_crit_hyst_store(), data is retrieved via dev_get_drvdata(dev) without calling lm63_update_device() to populate the cache. If a user writes to the sysfs node before any sensor read operation has occurred, data->temp8[2] will be 0. Does this cause the driver to calculate the relative hysteresis based on a critical temperature of 0 degrees instead of the real thermal limit, corrupting the chip's thermal protection configuration when written to the hardware register?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
#3 lore raw

Re: [PATCH 2/2] hwmon: (lm63) Add locking to avoid TOCTOU

Quoted 5 lines · On Thu, Apr 16, 2026 at 7:44 PM <[email protected]> wrote:

Sashiko AI review found 3 potential issue(s):
- [High] Pre-existing Torn Read Race Condition in show_fan() causes massive fan speed spikes
- [High] Pre-existing Uninitialized Cache Read in temp2_crit_hyst_store() corrupts hardware thermal limits
- [Medium] Incomplete Fix: Unaddressed TOCTOU Races in temp2_crit_hyst_show() and show_lut_temp_hyst()

Thanks for the detailed review.

I will add the missing locks for show_fan() and the hysteresis show functions in v2.

The potential uninitialized cache read in temp2_crit_hyst_store() needs further investigation. If it is a bug, I will address it in a separate patch, as it is a pre-existing issue unrelated to this series.

Quoted 43 lines

--

commit 1233dd514546736dfbbf09561bf360360c5933ca
Author: Gui-Dong Han <[email protected]>

hwmon: (lm63) Add locking to avoid TOCTOU

This commit adds locking to show_pwm1() and show_temp11() to prevent TOCTOU races when reading multiple cached values. It also states that other functions in the driver were checked for similar issues.

Quoted 3 lines

Check the other functions in the driver as well. Keep them unchanged because they either do not access shared cached values multiple times or already do so under lock.

Is this description accurate?

Looking at temp2_crit_hyst_show(), it calculates its value by reading both data->temp8[2] and data->temp2_crit_hyst. Similarly, show_lut_temp_hyst() reads data->temp8[attr->index] and data->lut_temp_hyst. Because neither function holds data->update_lock, can a concurrent cache update cause them to fetch a mixed snapshot (e.g., a newly updated base temperature but a stale hysteresis offset), resulting in corrupted absolute hysteresis values returned to userspace?

Also, looking at show_fan(), it accesses data->fan[attr->index] without holding data->update_lock. In lm63_update_device(), data->fan[0] is populated via two separate non-atomic assignments. Can show_fan() read the intermediate torn state where the MSB is still 0, causing the calculation (5400000 / reg) to report massive, incorrect fan speed spikes?

Finally, in temp2_crit_hyst_store(), data is retrieved via dev_get_drvdata(dev) without calling lm63_update_device() to populate the cache. If a user writes to the sysfs node before any sensor read operation has occurred, data->temp8[2] will be 0. Does this cause the driver to calculate the relative hysteresis based on a critical temperature of 0 degrees instead of the real thermal limit, corrupting the chip's thermal protection configuration when written to the hardware register?

--
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2

#4 lore raw
In reply to Gui-Dong Han
Sashiko AI review found 2 potential issue(s): - [High] Incomplete TOCTOU fix allows compiler-induced divide-by-zero crash in `show_fan()`. - [Medium] Lockless sysfs read in `show_fan()` causes torn…

Sashiko AI review found 2 potential issue(s):
- [High] Incomplete TOCTOU fix allows compiler-induced divide-by-zero crash in show_fan().
- [Medium] Lockless sysfs read in show_fan() causes torn reads and violates hwmon subsystem locking guidelines. --

commit ebc10731137639981842ea19f80f0b969db71842
Author: Gui-Dong Han <[email protected]>

hwmon: (lm63) Convert macro to function to avoid TOCTOU

The patch converts the FAN_FROM_REG macro to a static function fan_from_reg() to evaluate its argument only once, avoiding a TOCTOU race and potential divide-by-zero error.

Quoted 18 lines
drivers/hwmon/lm63.c +8 -2
index 035176a98ce9c..da132b267c584 100644
@@ -109,8 +109,14 @@ static const unsigned short normal_i2c[] = { 0x18, 0x4c, 0x4e, I2C_CLIENT_END };
  * adapted accordingly.
  */
 
-#define FAN_FROM_REG(reg)	((reg) == 0xFFFC || (reg) == 0 ? 0 : \
-				 5400000 / (reg))
+static int fan_from_reg(int reg)
+{
+	if (reg == 0xFFFC || reg == 0)
+		return 0;
+
+	return 5400000 / reg;
+}
+

[ ... ]

drivers/hwmon/lm63.c static ssize_t show_fan(struct device *dev, struct device_attribute *devattr, · line 342

Quoted 6 lines
drivers/hwmon/lm63.c +1 -1
@@ -333,7 +339,7 @@ static ssize_t show_fan(struct device *dev, struct device_attribute *devattr,
 {
 	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
 	struct lm63_data *data = lm63_update_device(dev);
-	return sprintf(buf, "%d\n", FAN_FROM_REG(data->fan[attr->index]));
+	return sprintf(buf, "%d\n", fan_from_reg(data->fan[attr->index]));

Since fan_from_reg() is a small static function, won't the compiler likely inline it?

If it gets inlined, couldn't the compiler optimize the code by reloading data->fan[attr->index] from memory multiple times (once for the 0 check, and again for the division)?

If lm63_update_device() concurrently writes to data->fan[0] in two non-atomic steps (where the first step can momentarily set the value to 0), and a reload happens exactly when the value is 0, could this still cause a divide-by-zero panic?

Note that the subsequent patch in the series ('hwmon: (lm63) Add locking to avoid TOCTOU') explicitly skips show_fan(), leaving this unfixed.

Additionally, is it safe to read data->fan locklessly here?

According to the hwmon subsystem guidelines, drivers must implement locking for attributes registered outside the core-serialized info API. Because this driver registers its attributes using devm_hwmon_device_register_with_groups(), it should lock its attribute accesses.

Since lm63_update_device() releases the update_lock before returning, and data->fan[0] is updated non-atomically in two steps, couldn't a concurrent sysfs read observe a partially updated 16-bit state?

Could this torn read result in wildly inaccurate fan speed values being reported? The subsequent patch ('hwmon: (lm63) Add locking to avoid TOCTOU') fixes similar locking issues but explicitly omits show_fan().

Quoted 1 line
drivers/hwmon/lm63.c +0 -0
 }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#5 lore raw
Quoted 48 lines · On Thu, Apr 16, 2026 at 6:41 PM <[email protected]> wrote:

Sashiko AI review found 2 potential issue(s):
- [High] Incomplete TOCTOU fix allows compiler-induced divide-by-zero crash in show_fan().
- [Medium] Lockless sysfs read in show_fan() causes torn reads and violates hwmon subsystem locking guidelines. --

commit ebc10731137639981842ea19f80f0b969db71842
Author: Gui-Dong Han <[email protected]>

hwmon: (lm63) Convert macro to function to avoid TOCTOU

The patch converts the FAN_FROM_REG macro to a static function fan_from_reg() to evaluate its argument only once, avoiding a TOCTOU race and potential divide-by-zero error.

Quoted 18 lines
drivers/hwmon/lm63.c +8 -2
index 035176a98ce9c..da132b267c584 100644
@@ -109,8 +109,14 @@ static const unsigned short normal_i2c[] = { 0x18, 0x4c, 0x4e, I2C_CLIENT_END };
  * adapted accordingly.
  */
 
-#define FAN_FROM_REG(reg)    ((reg) == 0xFFFC || (reg) == 0 ? 0 : \
-                              5400000 / (reg))
+static int fan_from_reg(int reg)
+{
+     if (reg == 0xFFFC || reg == 0)
+             return 0;
+
+     return 5400000 / reg;
+}
+

[ ... ]

drivers/hwmon/lm63.c static ssize_t show_fan(struct device *dev, struct device_attribute *devattr, · line 342

Quoted 6 lines
drivers/hwmon/lm63.c +1 -1
@@ -333,7 +339,7 @@ static ssize_t show_fan(struct device *dev, struct device_attribute *devattr,
 {
      struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
      struct lm63_data *data = lm63_update_device(dev);
-     return sprintf(buf, "%d\n", FAN_FROM_REG(data->fan[attr->index]));
+     return sprintf(buf, "%d\n", fan_from_reg(data->fan[attr->index]));

Since fan_from_reg() is a small static function, won't the compiler likely inline it?

Quoted 3 lines

If it gets inlined, couldn't the compiler optimize the code by reloading data->fan[attr->index] from memory multiple times (once for the 0 check, and again for the division)?

I am not an expert on compiler optimization. I consulted a few different models and got mixed answers. Unlike a macro with a side effect (like ++), an inlined function doesn't strictly force multiple memory fetches, so it's a bit ambiguous.

Quoted 11 lines

If lm63_update_device() concurrently writes to data->fan[0] in two non-atomic steps (where the first step can momentarily set the value to 0), and a reload happens exactly when the value is 0, could this still cause a divide-by-zero panic?

Note that the subsequent patch in the series ('hwmon: (lm63) Add locking to avoid TOCTOU') explicitly skips show_fan(), leaving this unfixed.

Additionally, is it safe to read data->fan locklessly here?

Quoted 8 lines

According to the hwmon subsystem guidelines, drivers must implement locking for attributes registered outside the core-serialized info API. Because this driver registers its attributes using devm_hwmon_device_register_with_groups(), it should lock its attribute accesses.

Since lm63_update_device() releases the update_lock before returning, and data->fan[0] is updated non-atomically in two steps, couldn't a concurrent sysfs read observe a partially updated 16-bit state?

Your point about data->fan[0] being updated non-atomically in two steps is a very nice finding. This gives me a solid reason to just add locking to show_fan() and fix the torn read properly.

Quoted 9 lines

Could this torn read result in wildly inaccurate fan speed values being reported? The subsequent patch ('hwmon: (lm63) Add locking to avoid TOCTOU') fixes similar locking issues but explicitly omits show_fan().

drivers/hwmon/lm63.c static ssize_t show_fan(struct device *dev, struct device_attribute *devattr, · line 343

Quoted 1 line
drivers/hwmon/lm63.c +0 -0
 }

--
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1