Re: [PATCH v1 1/2] Revert "thermal: hwmon: Use extra_groups for adding temperature attributes"

Thank you for your contribution! Sashiko AI review found 1 potential issue(=
s) to consider:
- [Medium] Adding sysfs attributes via `device_create_file()` after device =
registration creates a race condition with userspace uevents.
--

--- Patch [1]: [PATCH v1 1/2] Revert "thermal: hwmon: Use extra_groups for =
adding temperature attributes" ---
commit fd1537461eb002e63315e0164251174bf300f7cc
Author: Rafael J. Wysocki <[email protected]>

Revert "thermal: hwmon: Use extra_groups for adding temperature attributes"

This reverts commit cfb5dc0f60fb because it was depended on by another comm=
it
that turned out to be problematic.

> diff --git a/drivers/thermal/thermal_hwmon.c b/drivers/thermal/thermal_hw=
mon.c
> index 386dfb9f559e3..223ae1571655b 100644
> --- a/drivers/thermal/thermal_hwmon.c
> +++ b/drivers/thermal/thermal_hwmon.c
[ ... ]
> @@ -120,13 +114,35 @@ int thermal_add_hwmon_sysfs(struct thermal_zone_dev=
ice *tz)
>  	scnprintf(hwmon->name, THERMAL_HWMON_NAME_LENGTH, "%s_%d", tz->type, tz=
->id);
>  	strreplace(hwmon->name, '-', '_');
>  	hwmon->device =3D hwmon_device_register_for_thermal(&tz->device,
> -							  hwmon->name, hwmon,
> -							  thermal_hwmon_groups);
> +							  hwmon->name, hwmon);
>  	if (IS_ERR(hwmon->device)) {
> -		int result =3D PTR_ERR(hwmon->device);
> +		result =3D PTR_ERR(hwmon->device);
> +		goto free_mem;
> +	}
> =20
> -		kfree(hwmon);
> -		return result;
> +	temp =3D &hwmon->tz_temp;
> +
> +	temp->tz =3D tz;
> +
> +	temp->temp_input.attr.attr.name =3D "temp1_input";
> +	temp->temp_input.attr.attr.mode =3D 0444;
> +	temp->temp_input.attr.show =3D temp_input_show;
> +	sysfs_attr_init(&temp->temp_input.attr.attr);
> +	result =3D device_create_file(hwmon->device, &temp->temp_input.attr);

[Severity: Medium]
Does this sequence create a race condition with userspace uevents?=20

When thermal_add_hwmon_sysfs() calls hwmon_device_register_for_thermal(),
the device is registered and the KOBJ_ADD uevent fires immediately.

If userspace monitoring daemons or udev rules respond to this uevent and=20
read the hwmon device directory right away, they might query it before=20
device_create_file() finishes adding the temp1_input and temp1_crit=20
attributes.

Could this lead to missing sensor attributes in userspace during device
initialization?

--=20
Sashiko AI review =C2=B7 https://sashiko.dev/#/patchset/6319276.lOV4Wx5bFT@=
rafael.j.wysocki?part=3D1