]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
hwmon: (isl28022) Fix integer overflow in power calculation on 32-bit
authorSanman Pradhan <psanman@juniper.net>
Fri, 10 Apr 2026 00:26:19 +0000 (00:26 +0000)
committerGuenter Roeck <linux@roeck-us.net>
Fri, 10 Apr 2026 15:32:02 +0000 (08:32 -0700)
isl28022_read_power() computes:

  *val = ((51200000L * ((long)data->gain)) /
          (long)data->shunt) * (long)regval;

On 32-bit platforms, 'long' is 32 bits. With gain=8 and shunt=10000
(the default configuration):

  (51200000 * 8) / 10000 = 40960
  40960 * 65535 = 2,684,313,600

This exceeds LONG_MAX (2,147,483,647), resulting in signed integer
overflow.

Additionally, dividing before multiplying by regval loses precision
unnecessarily.

Use u64 arithmetic with div_u64() and multiply before dividing to
retain precision. The intermediate product cannot overflow u64
(worst case: 51200000 * 8 * 65535 = 26843136000000). Power is
inherently non-negative, so unsigned types are the natural fit.
Cap the result to LONG_MAX before returning it through the hwmon
callback.

Fixes: 39671a14df4f2 ("hwmon: (isl28022) new driver for ISL28022 power monitor")
Cc: stable@vger.kernel.org
Signed-off-by: Sanman Pradhan <psanman@juniper.net>
Link: https://lore.kernel.org/r/20260410002613.424557-1-sanman.pradhan@hpe.com
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
drivers/hwmon/isl28022.c

index c2e559dde63f6188361f83b0cf736689de92d4db..c5a34ceedcdb2832602e297c0a93b60009f18a32 100644 (file)
@@ -9,6 +9,7 @@
 #include <linux/err.h>
 #include <linux/hwmon.h>
 #include <linux/i2c.h>
+#include <linux/math64.h>
 #include <linux/module.h>
 #include <linux/regmap.h>
 
@@ -185,8 +186,8 @@ static int isl28022_read_power(struct device *dev, u32 attr, long *val)
                                  ISL28022_REG_POWER, &regval);
                if (err < 0)
                        return err;
-               *val = ((51200000L * ((long)data->gain)) /
-                       (long)data->shunt) * (long)regval;
+               *val = min(div_u64(51200000ULL * data->gain * regval,
+                                  data->shunt), LONG_MAX);
                break;
        default:
                return -EOPNOTSUPP;