[PATCH v2 0/2] ACPI: TAD: Use auto-cleanup macros for runtime PM

[PATCH v2 0/2] ACPI: TAD: Use auto-cleanup macros for runtime PM

[Rafael J. Wysocki]

Hi All,

This supersedes

https://lore.kernel.org/linux-pm/3925484.kQq0lBPeGt@rafael.j.wysocki/

the first patch of which proved to be controversial (quite as expected), so it
has been dropped and the patch that was the second one previously is [1/2] now,
and it has not been changed.

Patch [2/2] (previously [3/3]) has been updated to use the full ACQUIRE()/
ACQUIRE_ERR() syntax.

Thanks!

[PATCH v2 1/2] ACPI: TAD: Rearrange runtime PM operations in acpi_tad_remove()

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

It is not necessary to resume the device upfront in acpi_tad_remove()
because both acpi_tad_disable_timer() and acpi_tad_clear_status()
attempt to resume it, but it is better to prevent it from suspending
between these calls by incrementing its runtime PM usage counter.

Accordingly, replace the pm_runtime_get_sync() call in acpi_tad_remove()
with a pm_runtime_get_noresume() one and put the latter right before the
first invocation of acpi_tad_disable_timer().

In addition, use pm_runtime_put_noidle() to drop the device's runtime
PM usage counter after using pm_runtime_get_noresume() to bump it up
to follow a common pattern and use pm_runtime_suspend() for suspending
the device afterward.

No intentional functional impact.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
---
 drivers/acpi/acpi_tad.c |    7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

--- a/drivers/acpi/acpi_tad.c
+++ b/drivers/acpi/acpi_tad.c
@@ -563,8 +563,6 @@ static void acpi_tad_remove(struct platf
 
 	device_init_wakeup(dev, false);
 
-	pm_runtime_get_sync(dev);
-
 	if (dd->capabilities & ACPI_TAD_RT)
 		sysfs_remove_group(&dev->kobj, &acpi_tad_time_attr_group);
 
@@ -573,6 +571,8 @@ static void acpi_tad_remove(struct platf
 
 	sysfs_remove_group(&dev->kobj, &acpi_tad_attr_group);
 
+	pm_runtime_get_noresume(dev);
+
 	acpi_tad_disable_timer(dev, ACPI_TAD_AC_TIMER);
 	acpi_tad_clear_status(dev, ACPI_TAD_AC_TIMER);
 	if (dd->capabilities & ACPI_TAD_DC_WAKE) {
@@ -580,7 +580,8 @@ static void acpi_tad_remove(struct platf
 		acpi_tad_clear_status(dev, ACPI_TAD_DC_TIMER);
 	}
 
-	pm_runtime_put_sync(dev);
+	pm_runtime_put_noidle(dev);
+	pm_runtime_suspend(dev);
 	pm_runtime_disable(dev);
 	acpi_remove_cmos_rtc_space_handler(handle);
 }

[PATCH v2 2/2] ACPI: TAD: Improve runtime PM using guard macros

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Use guard pm_runtime_active_try to simplify runtime PM cleanup and
implement runtime resume error handling in multiple places.

Also use guard pm_runtime_noresume to simplify acpi_tad_remove().

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
---
 drivers/acpi/acpi_tad.c |   57 +++++++++++++++++++++---------------------------
 1 file changed, 25 insertions(+), 32 deletions(-)

--- a/drivers/acpi/acpi_tad.c
+++ b/drivers/acpi/acpi_tad.c
@@ -90,12 +90,11 @@ static int acpi_tad_set_real_time(struct
 	args[0].buffer.pointer = (u8 *)rt;
 	args[0].buffer.length = sizeof(*rt);
 
-	pm_runtime_get_sync(dev);
+	ACQUIRE(pm_runtime_active_try, pm)(dev);
+	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
+		return -ENXIO;
 
 	status = acpi_evaluate_integer(handle, "_SRT", &arg_list, &retval);
-
-	pm_runtime_put_sync(dev);
-
 	if (ACPI_FAILURE(status) || retval)
 		return -EIO;
 
@@ -111,12 +110,11 @@ static int acpi_tad_get_real_time(struct
 	acpi_status status;
 	int ret = -EIO;
 
-	pm_runtime_get_sync(dev);
+	ACQUIRE(pm_runtime_active_try, pm)(dev);
+	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
+		return -ENXIO;
 
 	status = acpi_evaluate_object(handle, "_GRT", NULL, &output);
-
-	pm_runtime_put_sync(dev);
-
 	if (ACPI_FAILURE(status))
 		goto out_free;
 
@@ -266,12 +264,11 @@ static int acpi_tad_wake_set(struct devi
 	args[0].integer.value = timer_id;
 	args[1].integer.value = value;
 
-	pm_runtime_get_sync(dev);
+	ACQUIRE(pm_runtime_active_try, pm)(dev);
+	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
+		return -ENXIO;
 
 	status = acpi_evaluate_integer(handle, method, &arg_list, &retval);
-
-	pm_runtime_put_sync(dev);
-
 	if (ACPI_FAILURE(status) || retval)
 		return -EIO;
 
@@ -314,12 +311,11 @@ static ssize_t acpi_tad_wake_read(struct
 
 	args[0].integer.value = timer_id;
 
-	pm_runtime_get_sync(dev);
+	ACQUIRE(pm_runtime_active_try, pm)(dev);
+	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
+		return -ENXIO;
 
 	status = acpi_evaluate_integer(handle, method, &arg_list, &retval);
-
-	pm_runtime_put_sync(dev);
-
 	if (ACPI_FAILURE(status))
 		return -EIO;
 
@@ -370,12 +366,11 @@ static int acpi_tad_clear_status(struct
 
 	args[0].integer.value = timer_id;
 
-	pm_runtime_get_sync(dev);
+	ACQUIRE(pm_runtime_active_try, pm)(dev);
+	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
+		return -ENXIO;
 
 	status = acpi_evaluate_integer(handle, "_CWS", &arg_list, &retval);
-
-	pm_runtime_put_sync(dev);
-
 	if (ACPI_FAILURE(status) || retval)
 		return -EIO;
 
@@ -411,12 +406,11 @@ static ssize_t acpi_tad_status_read(stru
 
 	args[0].integer.value = timer_id;
 
-	pm_runtime_get_sync(dev);
+	ACQUIRE(pm_runtime_active_try, pm)(dev);
+	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
+		return -ENXIO;
 
 	status = acpi_evaluate_integer(handle, "_GWS", &arg_list, &retval);
-
-	pm_runtime_put_sync(dev);
-
 	if (ACPI_FAILURE(status))
 		return -EIO;
 
@@ -571,16 +565,15 @@ static void acpi_tad_remove(struct platf
 
 	sysfs_remove_group(&dev->kobj, &acpi_tad_attr_group);
 
-	pm_runtime_get_noresume(dev);
-
-	acpi_tad_disable_timer(dev, ACPI_TAD_AC_TIMER);
-	acpi_tad_clear_status(dev, ACPI_TAD_AC_TIMER);
-	if (dd->capabilities & ACPI_TAD_DC_WAKE) {
-		acpi_tad_disable_timer(dev, ACPI_TAD_DC_TIMER);
-		acpi_tad_clear_status(dev, ACPI_TAD_DC_TIMER);
+	scoped_guard(pm_runtime_noresume, dev) {
+		acpi_tad_disable_timer(dev, ACPI_TAD_AC_TIMER);
+		acpi_tad_clear_status(dev, ACPI_TAD_AC_TIMER);
+		if (dd->capabilities & ACPI_TAD_DC_WAKE) {
+			acpi_tad_disable_timer(dev, ACPI_TAD_DC_TIMER);
+			acpi_tad_clear_status(dev, ACPI_TAD_DC_TIMER);
+		}
 	}
 
-	pm_runtime_put_noidle(dev);
 	pm_runtime_suspend(dev);
 	pm_runtime_disable(dev);
 	acpi_remove_cmos_rtc_space_handler(handle);

Re: [PATCH v2 2/2] ACPI: TAD: Improve runtime PM using guard macros

[Jonathan Cameron]

On Sat, 18 Oct 2025 14:24:42 +0200
"Rafael J. Wysocki" <rafael@kernel.org> wrote:

> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> 
> Use guard pm_runtime_active_try to simplify runtime PM cleanup and
> implement runtime resume error handling in multiple places.
> 
> Also use guard pm_runtime_noresume to simplify acpi_tad_remove().
> 
> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> ---
>  drivers/acpi/acpi_tad.c |   57 +++++++++++++++++++++---------------------------
>  1 file changed, 25 insertions(+), 32 deletions(-)
> 
> --- a/drivers/acpi/acpi_tad.c
> +++ b/drivers/acpi/acpi_tad.c

> @@ -111,12 +110,11 @@ static int acpi_tad_get_real_time(struct
>  	acpi_status status;
>  	int ret = -EIO;
>  
> -	pm_runtime_get_sync(dev);
> +	ACQUIRE(pm_runtime_active_try, pm)(dev);
> +	if (ACQUIRE_ERR(pm_runtime_active_try, &pm))
> +		return -ENXIO;
>  
>  	status = acpi_evaluate_object(handle, "_GRT", NULL, &output);
> -
> -	pm_runtime_put_sync(dev);
> -
>  	if (ACPI_FAILURE(status))

Whilst it isn't actually a bug, this does run up against the guidance
in cleanup.h to avoid mixing gotos and cleanup.h usage in a single function.
That's partly a simplification to avoid having to explain the issues with
jumping past inline declarations.

If you want to follow that guidance, either you'd need to add a helper along the
lines of:

DEFINE_FREE(acpi_buffer, void *m ACPI_FREE(_T));
void *acpi_eval_grt(acpi_handle handle, struct acpi_buffer *output)
{
	acpi_status status = acpi_evaluate_object(handle, "_GRT", NULL, &output);

	if (ACPI_FAILURE(status)) {
		ACPI_FREE(output.pointer);
		return ERR_PTR(-EIO); //whatever error makse sense.
	}

	return output.pointer;
}

void *out_obj __free(acpi_buffer) = acpi_eval_grt(handle, &output);

Then can return directly at all error paths.  Not the nicest bit of
code though.

Or, factor out everthing after that allocation down
to the label as helper function and have direct returns in that.

or leave it all as is and hope Linus doesn't get grumpy about mix
and match (which lead to that guidance being so general in the first place!)

The rest look good to me.

Jonathan



>  		goto out_free;
>