[v1,2/2] thermal: sysfs: Eliminate unnecessary zone locking

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

The _show() callback functions of the trip point sysfs attributes,
temperature, hysteresis and type, need not use thermal zone locking,
because the layout of the data structures they access does not change
after the thermal zone registration.

Namely, they all need to access a specific entry in the thermal
zone's trips[] table that is always present for non-tripless thermal
zones and its size cannot change after the thermal zone has been
registered.  Thus it is always safe to access the trips[] table of a
registered thermal zone from each of the sysfs attributes in question.

Moreover, the type of a trip point does not change after registering its
thermal zone, and while its temperature and hysteresis can change, for
example due to a firmware-induced thermal zone update, holding the zone
lock around reading them is pointless, because it does not prevent stale
values from being returned to user space.  For example, a trip point
temperature can always change ater trip_point_temp_show() has read it
and before the function's return statement is executed, regardless of
whether or not zone locking is used.

For this reason, drop the zone locking from trip_point_type_show(),
trip_point_temp_show(), and trip_point_hyst_show().

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
---
 drivers/thermal/thermal_sysfs.c |   60 ++++++++++++++--------------------------
 1 file changed, 21 insertions(+), 39 deletions(-)
  

On Thu, Nov 30, 2023 at 7:38 PM Rafael J. Wysocki <rjw@rjwysocki.net> wrote:
>
> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>
> The _show() callback functions of the trip point sysfs attributes,
> temperature, hysteresis and type, need not use thermal zone locking,
> because the layout of the data structures they access does not change
> after the thermal zone registration.
>
> Namely, they all need to access a specific entry in the thermal
> zone's trips[] table that is always present for non-tripless thermal
> zones and its size cannot change after the thermal zone has been
> registered.  Thus it is always safe to access the trips[] table of a
> registered thermal zone from each of the sysfs attributes in question.
>
> Moreover, the type of a trip point does not change after registering its
> thermal zone, and while its temperature and hysteresis can change, for
> example due to a firmware-induced thermal zone update, holding the zone
> lock around reading them is pointless, because it does not prevent stale
> values from being returned to user space.  For example, a trip point
> temperature can always change ater trip_point_temp_show() has read it
> and before the function's return statement is executed, regardless of
> whether or not zone locking is used.
>
> For this reason, drop the zone locking from trip_point_type_show(),
> trip_point_temp_show(), and trip_point_hyst_show().
>
> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> ---
>  drivers/thermal/thermal_sysfs.c |   60 ++++++++++++++--------------------------
>  1 file changed, 21 insertions(+), 39 deletions(-)
>
> Index: linux-pm/drivers/thermal/thermal_sysfs.c
> ===================================================================
> --- linux-pm.orig/drivers/thermal/thermal_sysfs.c
> +++ linux-pm/drivers/thermal/thermal_sysfs.c
> @@ -83,25 +83,18 @@ trip_point_type_show(struct device *dev,
>                      char *buf)
>  {
>         struct thermal_zone_device *tz = to_thermal_zone(dev);
> -       struct thermal_trip trip;
> -       int trip_id, result;
> +       int trip_id;
> +
> +       if (!device_is_registered(dev))
> +               return -ENODEV;
>
>         if (sscanf(attr->attr.name, "trip_point_%d_type", &trip_id) != 1)
>                 return -EINVAL;
>
> -       mutex_lock(&tz->lock);
> -
> -       if (device_is_registered(dev))
> -               result = __thermal_zone_get_trip(tz, trip_id, &trip);
> -       else
> -               result = -ENODEV;
> -
> -       mutex_unlock(&tz->lock);
> -
> -       if (result)
> -               return result;
> +       if (trip_id < 0 || trip_id > tz->num_trips)

An off-by-one here, it should be trip_id >= tz->num_trips (and
analogously below).

I'll send an update of this shortly.

> +               return -EINVAL;
>
> -       switch (trip.type) {
> +       switch (tz->trips[trip_id].type) {
>         case THERMAL_TRIP_CRITICAL:
>                 return sprintf(buf, "critical\n");
>         case THERMAL_TRIP_HOT:
> @@ -164,25 +157,18 @@ trip_point_temp_show(struct device *dev,
>                      char *buf)
>  {
>         struct thermal_zone_device *tz = to_thermal_zone(dev);
> -       struct thermal_trip trip;
> -       int trip_id, ret;
> +       int trip_id;
> +
> +       if (!device_is_registered(dev))
> +               return -ENODEV;
>
>         if (sscanf(attr->attr.name, "trip_point_%d_temp", &trip_id) != 1)
>                 return -EINVAL;
>
> -       mutex_lock(&tz->lock);
> -
> -       if (device_is_registered(dev))
> -               ret = __thermal_zone_get_trip(tz, trip_id, &trip);
> -       else
> -               ret = -ENODEV;
> -
> -       mutex_unlock(&tz->lock);
> -
> -       if (ret)
> -               return ret;
> +       if (trip_id < 0 || trip_id > tz->num_trips)
> +               return -EINVAL;
>
> -       return sprintf(buf, "%d\n", trip.temperature);
> +       return sprintf(buf, "%d\n", tz->trips[trip_id].temperature);
>  }
>
>  static ssize_t
> @@ -234,22 +220,18 @@ trip_point_hyst_show(struct device *dev,
>                      char *buf)
>  {
>         struct thermal_zone_device *tz = to_thermal_zone(dev);
> -       struct thermal_trip trip;
> -       int trip_id, ret;
> +       int trip_id;
> +
> +       if (!device_is_registered(dev))
> +               return -ENODEV;
>
>         if (sscanf(attr->attr.name, "trip_point_%d_hyst", &trip_id) != 1)
>                 return -EINVAL;
>
> -       mutex_lock(&tz->lock);
> -
> -       if (device_is_registered(dev))
> -               ret = __thermal_zone_get_trip(tz, trip_id, &trip);
> -       else
> -               ret = -ENODEV;
> -
> -       mutex_unlock(&tz->lock);
> +       if (trip_id < 0 || trip_id > tz->num_trips)
> +               return -EINVAL;
>
> -       return ret ? ret : sprintf(buf, "%d\n", trip.hysteresis);
> +       return sprintf(buf, "%d\n", tz->trips[trip_id].hysteresis);
>  }
>
>  static ssize_t
>
>
>
>
  

On 30/11/2023 19:37, Rafael J. Wysocki wrote:
> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> 
> The _show() callback functions of the trip point sysfs attributes,
> temperature, hysteresis and type, need not use thermal zone locking,
> because the layout of the data structures they access does not change
> after the thermal zone registration.
> 
> Namely, they all need to access a specific entry in the thermal
> zone's trips[] table that is always present for non-tripless thermal
> zones and its size cannot change after the thermal zone has been
> registered.  Thus it is always safe to access the trips[] table of a
> registered thermal zone from each of the sysfs attributes in question.
> 
> Moreover, the type of a trip point does not change after registering its
> thermal zone, and while its temperature and hysteresis can change, for
> example due to a firmware-induced thermal zone update, holding the zone
> lock around reading them is pointless, because it does not prevent stale
> values from being returned to user space.  For example, a trip point
> temperature can always change ater trip_point_temp_show() has read it
> and before the function's return statement is executed, regardless of
> whether or not zone locking is used.
> 
> For this reason, drop the zone locking from trip_point_type_show(),
> trip_point_temp_show(), and trip_point_hyst_show().

Isn't the lock used to protect against device_del() from 
thermal_zone_device_unregister() ?


> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> ---
>   drivers/thermal/thermal_sysfs.c |   60 ++++++++++++++--------------------------
>   1 file changed, 21 insertions(+), 39 deletions(-)
> 
> Index: linux-pm/drivers/thermal/thermal_sysfs.c
> ===================================================================
> --- linux-pm.orig/drivers/thermal/thermal_sysfs.c
> +++ linux-pm/drivers/thermal/thermal_sysfs.c
> @@ -83,25 +83,18 @@ trip_point_type_show(struct device *dev,
>   		     char *buf)
>   {
>   	struct thermal_zone_device *tz = to_thermal_zone(dev);
> -	struct thermal_trip trip;
> -	int trip_id, result;
> +	int trip_id;
> +
> +	if (!device_is_registered(dev))
> +		return -ENODEV;
>   
>   	if (sscanf(attr->attr.name, "trip_point_%d_type", &trip_id) != 1)
>   		return -EINVAL;
>   
> -	mutex_lock(&tz->lock);
> -
> -	if (device_is_registered(dev))
> -		result = __thermal_zone_get_trip(tz, trip_id, &trip);
> -	else
> -		result = -ENODEV;
> -
> -	mutex_unlock(&tz->lock);
> -
> -	if (result)
> -		return result;
> +	if (trip_id < 0 || trip_id > tz->num_trips)
> +		return -EINVAL;
>   
> -	switch (trip.type) {
> +	switch (tz->trips[trip_id].type) {
>   	case THERMAL_TRIP_CRITICAL:
>   		return sprintf(buf, "critical\n");
>   	case THERMAL_TRIP_HOT:
> @@ -164,25 +157,18 @@ trip_point_temp_show(struct device *dev,
>   		     char *buf)
>   {
>   	struct thermal_zone_device *tz = to_thermal_zone(dev);
> -	struct thermal_trip trip;
> -	int trip_id, ret;
> +	int trip_id;
> +
> +	if (!device_is_registered(dev))
> +		return -ENODEV;
>   
>   	if (sscanf(attr->attr.name, "trip_point_%d_temp", &trip_id) != 1)
>   		return -EINVAL;
>   
> -	mutex_lock(&tz->lock);
> -
> -	if (device_is_registered(dev))
> -		ret = __thermal_zone_get_trip(tz, trip_id, &trip);
> -	else
> -		ret = -ENODEV;
> -
> -	mutex_unlock(&tz->lock);
> -
> -	if (ret)
> -		return ret;
> +	if (trip_id < 0 || trip_id > tz->num_trips)
> +		return -EINVAL;
>   
> -	return sprintf(buf, "%d\n", trip.temperature);
> +	return sprintf(buf, "%d\n", tz->trips[trip_id].temperature);
>   }
>   
>   static ssize_t
> @@ -234,22 +220,18 @@ trip_point_hyst_show(struct device *dev,
>   		     char *buf)
>   {
>   	struct thermal_zone_device *tz = to_thermal_zone(dev);
> -	struct thermal_trip trip;
> -	int trip_id, ret;
> +	int trip_id;
> +
> +	if (!device_is_registered(dev))
> +		return -ENODEV;
>   
>   	if (sscanf(attr->attr.name, "trip_point_%d_hyst", &trip_id) != 1)
>   		return -EINVAL;
>   
> -	mutex_lock(&tz->lock);
> -
> -	if (device_is_registered(dev))
> -		ret = __thermal_zone_get_trip(tz, trip_id, &trip);
> -	else
> -		ret = -ENODEV;
> -
> -	mutex_unlock(&tz->lock);
> +	if (trip_id < 0 || trip_id > tz->num_trips)
> +		return -EINVAL;
>   
> -	return ret ? ret : sprintf(buf, "%d\n", trip.hysteresis);
> +	return sprintf(buf, "%d\n", tz->trips[trip_id].hysteresis);
>   }
>   
>   static ssize_t
> 
> 
> 
>
  

On Fri, Dec 1, 2023 at 10:16 AM Daniel Lezcano
<daniel.lezcano@linaro.org> wrote:
>
> On 30/11/2023 19:37, Rafael J. Wysocki wrote:
> > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> >
> > The _show() callback functions of the trip point sysfs attributes,
> > temperature, hysteresis and type, need not use thermal zone locking,
> > because the layout of the data structures they access does not change
> > after the thermal zone registration.
> >
> > Namely, they all need to access a specific entry in the thermal
> > zone's trips[] table that is always present for non-tripless thermal
> > zones and its size cannot change after the thermal zone has been
> > registered.  Thus it is always safe to access the trips[] table of a
> > registered thermal zone from each of the sysfs attributes in question.
> >
> > Moreover, the type of a trip point does not change after registering its
> > thermal zone, and while its temperature and hysteresis can change, for
> > example due to a firmware-induced thermal zone update, holding the zone
> > lock around reading them is pointless, because it does not prevent stale
> > values from being returned to user space.  For example, a trip point
> > temperature can always change ater trip_point_temp_show() has read it
> > and before the function's return statement is executed, regardless of
> > whether or not zone locking is used.
> >
> > For this reason, drop the zone locking from trip_point_type_show(),
> > trip_point_temp_show(), and trip_point_hyst_show().
>
> Isn't the lock used to protect against device_del() from
> thermal_zone_device_unregister() ?

Ah, I missed the zone locking around device_del() in there.

OK, please scratch this series for now.