Re: [PATCH 1/1] platform/x86: hp-wmi: fix platform profile issues on generic HP laptops

On Sat, Jun 20, 2026 at 12:09:11AM +0200, Marco Scardovi wrote:
>On generic (non-Omen/non-Victus) HP laptops supported by the
>hp-wmi driver, the platform_profile sysfs operations could fail,
>leading to 'platform_profile: Failed to get profile for handler hp-wmi'
>errors and preventing userspace power profile management from working
>correctly.
>
>The driver was blindly registering all 4 profiles (quiet, cool,
>balanced, performance) without checking which ones were actually
>supported by the BIOS. Furthermore, when userspace switched profiles,
>hp_wmi_platform_profile_get() was called, which queried the BIOS
>using thermal_profile_get(). If the BIOS query returned an error or
>returned a value not in the recognized cases, the get call failed,
>throwing an error.
>
>Fix this by:
>1. Dynamically probing which thermal profiles are supported by the BIOS
>   during driver registration by temporarily setting each profile and
>   checking the return code, then restoring the original profile.
>2. Initializing and updating active_platform_profile to cache the
>   last set/boot profile.
>3. Falling back to the cached active_platform_profile in
>   hp_wmi_platform_profile_get() if the BIOS query fails or returns
>   an invalid/unmapped value, rather than returning an error.
>
>Closes: https://bugzilla.kernel.org/show_bug.cgi?id=220008

Is the patch tested by them?

>Closes: https://bugzilla.kernel.org/show_bug.cgi?id=221569
>Closes: https://lore.kernel.org/platform-driver-x86/[email protected]/
>Fixes: 4296f679ca50 ("platform/x86: hp-wmi: add platform profile support")
>Assisted-by: Antigravity:gemini-3.5-flash
>Signed-off-by: Marco Scardovi <[email protected]>
>---
> drivers/platform/x86/hp/hp-wmi.c | 53 +++++++++++++++++++++++++++++---
> 1 file changed, 48 insertions(+), 5 deletions(-)
>
>diff --git a/drivers/platform/x86/hp/hp-wmi.c b/drivers/platform/x86/hp/hp-wmi.c
>index 8ba286ed8721..93df269445eb 100644
>--- a/drivers/platform/x86/hp/hp-wmi.c
>+++ b/drivers/platform/x86/hp/hp-wmi.c
>@@ -1662,8 +1662,11 @@ static int hp_wmi_platform_profile_get(struct device *dev,
> 	int tp;
>
> 	tp = thermal_profile_get();
>-	if (tp < 0)
>-		return tp;
>+	if (tp < 0) {
>+		guard(mutex)(&active_platform_profile_lock);
>+		*profile = active_platform_profile;
>+		return 0;
>+	}

This does silence the error but I don't think that's a wise decision
since it effectively makes userspace blind to any WMI failure. Ideally,
if thermal_profile_get() fails then the board isn't supposed to be using
HPWMI_THERMAL_PROFILE_QUERY.

>
> 	switch (tp) {
> 	case HP_THERMAL_PROFILE_PERFORMANCE:
>@@ -1679,7 +1682,9 @@ static int hp_wmi_platform_profile_get(struct device *dev,
> 		*profile = PLATFORM_PROFILE_QUIET;
> 		break;
> 	default:
>-		return -EINVAL;
>+		guard(mutex)(&active_platform_profile_lock);
>+		*profile = active_platform_profile;
>+		break;
> 	}
>
> 	return 0;
>@@ -1707,10 +1712,14 @@ static int hp_wmi_platform_profile_set(struct device *dev,
> 		return -EOPNOTSUPP;
> 	}
>
>+	guard(mutex)(&active_platform_profile_lock);
>+
> 	err = thermal_profile_set(tp);
> 	if (err)
> 		return err;
>
>+	active_platform_profile = profile;
>+
> 	return 0;
> }
>
>@@ -2017,8 +2026,23 @@ static int hp_wmi_platform_profile_probe(void *drvdata, unsigned long *choices)
> 		/* Adding an equivalent to HP Omen software ECO mode: */
> 		set_bit(PLATFORM_PROFILE_LOW_POWER, choices);
> 	} else {
>-		set_bit(PLATFORM_PROFILE_QUIET, choices);
>-		set_bit(PLATFORM_PROFILE_COOL, choices);
>+		int current_tp = thermal_profile_get();
>+
>+		if (current_tp < 0)
>+			return current_tp;
>+
>+		if (thermal_profile_set(HP_THERMAL_PROFILE_QUIET) == 0)
>+			set_bit(PLATFORM_PROFILE_QUIET, choices);
>+		if (thermal_profile_set(HP_THERMAL_PROFILE_COOL) == 0)
>+			set_bit(PLATFORM_PROFILE_COOL, choices);
>+		if (thermal_profile_set(HP_THERMAL_PROFILE_DEFAULT) == 0)
>+			set_bit(PLATFORM_PROFILE_BALANCED, choices);
>+		if (thermal_profile_set(HP_THERMAL_PROFILE_PERFORMANCE) == 0)
>+			set_bit(PLATFORM_PROFILE_PERFORMANCE, choices);

Probing for HP_THERMAL_PROFILE_DEFAULT and HP_THERMAL_PROFILE_PERFORMANCE
seem redundant since those bits are unconditionally set in this function
after the if-else blocks.

>+
>+		/* Restore the original thermal profile */
>+		thermal_profile_set(current_tp);
>+		return 0;
> 	}
>
> 	set_bit(PLATFORM_PROFILE_BALANCED, choices);
>@@ -2263,6 +2287,25 @@ static int thermal_profile_setup(struct platform_device *device)
> 		if (err)
> 			return err;
>
>+		/* Initialize active_platform_profile */
>+		switch (tp) {
>+		case HP_THERMAL_PROFILE_PERFORMANCE:
>+			active_platform_profile = PLATFORM_PROFILE_PERFORMANCE;
>+			break;
>+		case HP_THERMAL_PROFILE_DEFAULT:
>+			active_platform_profile = PLATFORM_PROFILE_BALANCED;
>+			break;
>+		case HP_THERMAL_PROFILE_COOL:
>+			active_platform_profile = PLATFORM_PROFILE_COOL;
>+			break;
>+		case HP_THERMAL_PROFILE_QUIET:
>+			active_platform_profile = PLATFORM_PROFILE_QUIET;
>+			break;
>+		default:
>+			active_platform_profile = PLATFORM_PROFILE_BALANCED;
>+			break;
>+		}
>+
> 		ops = &hp_wmi_platform_profile_ops;
> 	}
>
>-- 
>2.54.0
>