Re: [PATCH 1/5] hwmon: Export sensor and attribute definitions to userspace

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] hwmon: UAPI exposure of hwmon_max sentinel and internal bitmasks
--

commit bbc47a707f27865e8c59d7c4ea18a1b742278dfd
Author: Jihong Min <hurryman2212@gmail.com>

hwmon: Export sensor and attribute definitions to userspace

Move sensor types, attribute identifiers and configuration bits to
linux/hwmon.h in the UAPI so kernel and userspace drivers share them.
Keep existing values unchanged and use userspace-safe bit definitions.

> diff --git a/include/uapi/linux/hwmon.h b/include/uapi/linux/hwmon.h
> new file mode 100644
> index 0000000000000..274f9c9dcb87e
> --- /dev/null
> +++ b/include/uapi/linux/hwmon.h
> @@ -0,0 +1,372 @@
[ ... ]
> +enum hwmon_sensor_types {
> +	hwmon_chip = 0,
> +	hwmon_temp = 1,
> +	hwmon_in = 2,
> +	hwmon_curr = 3,
> +	hwmon_power = 4,
> +	hwmon_energy = 5,
> +	hwmon_energy64 = 6,
> +	hwmon_humidity = 7,
> +	hwmon_fan = 8,
> +	hwmon_pwm = 9,
> +	hwmon_intrusion = 10,
> +	hwmon_max = 11,
        ^^^^^^^^^^^^^^

[Severity: Medium]
Could exporting this expandable max boundary integer to the stable userspace 
UAPI header lead to out-of-bounds memory accesses in userspace applications?

If a userspace program statically sizes an array based on hwmon_max, and a 
future kernel release adds a new sensor type (incrementing hwmon_max), the 
program could index past the end of its array when receiving the new sensor 
type.

> +};
> +
> +enum hwmon_chip_attributes {
> +	hwmon_chip_temp_reset_history = 0,
> +	hwmon_chip_in_reset_history = 1,
> +	hwmon_chip_curr_reset_history = 2,
> +	hwmon_chip_power_reset_history = 3,
> +	hwmon_chip_register_tz = 4,
> +	hwmon_chip_update_interval = 5,
> +	hwmon_chip_update_interval_us = 6,
> +	hwmon_chip_alarms = 7,
> +	hwmon_chip_samples = 8,
> +	hwmon_chip_curr_samples = 9,
> +	hwmon_chip_in_samples = 10,
> +	hwmon_chip_power_samples = 11,
> +	hwmon_chip_temp_samples = 12,
> +	hwmon_chip_beep_enable = 13,
> +	hwmon_chip_pec = 14,
> +};
> +
> +#define HWMON_C_TEMP_RESET_HISTORY _BITUL(hwmon_chip_temp_reset_history)
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[Severity: Medium]
Is it necessary to export these configuration bitmasks to the UAPI?

These bitmasks are strictly used by the kernel-internal struct 
hwmon_channel_info and provide no value to userspace. Exporting them allows 
and encourages anti-patterns in userspace.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/cover.1791531025.git.hurryman2212@gmail.com?part=1