[PATCH] thermal: intel: intel_soc_dts_iosf: Fix thermal_zone removal

[PATCH] thermal: intel: intel_soc_dts_iosf: Fix thermal_zone removal

[Zhang Rui]

All of the existing callers of remove_dts_thermal_zone() pass a valid
pointer as the argument, so checking for the NULL pointer is redundant.

Plus, when calling remove_dts_thermal_zone() from
intel_soc_dts_iosf_init(), it is possible that
1. dts->tzone is an error pointer, when the sensor fails to be
   registered as a valid thermal zone
2. dts->tzone is unregistered in add_dts_thermal_zone(), when some
   failure occurs after thermal zone registered
In both cases, there is no need to unregister dts->tzone in
remove_dts_thermal_zone().

Clear dst->tzone when add_dts_thermal_zone() fails. And do thermal zone
removal in remove_dts_thermal_zone() only when dts->tzone is set.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
---
 drivers/thermal/intel/intel_soc_dts_iosf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/thermal/intel/intel_soc_dts_iosf.c b/drivers/thermal/intel/intel_soc_dts_iosf.c
index 7a66d0f077b0..c5203ba8f0b9 100644
--- a/drivers/thermal/intel/intel_soc_dts_iosf.c
+++ b/drivers/thermal/intel/intel_soc_dts_iosf.c
@@ -212,7 +212,7 @@ static int soc_dts_enable(int id)
 
 static void remove_dts_thermal_zone(struct intel_soc_dts_sensor_entry *dts)
 {
-	if (dts) {
+	if (dts->tzone) {
 		iosf_mbi_write(BT_MBI_UNIT_PMC, MBI_REG_WRITE,
 			       SOC_DTS_OFFSET_ENABLE, dts->store_status);
 		thermal_zone_device_unregister(dts->tzone);
@@ -277,6 +277,7 @@ static int add_dts_thermal_zone(int id, struct intel_soc_dts_sensor_entry *dts,
 err_enable:
 	thermal_zone_device_unregister(dts->tzone);
 err_ret:
+	dts->tzone = NULL;
 	return ret;
 }
 
-- 
2.34.1

Re: [PATCH] thermal: intel: intel_soc_dts_iosf: Fix thermal_zone removal

[Rafael J. Wysocki]

On Fri, Aug 11, 2023 at 10:45 AM Zhang Rui <rui.zhang@intel.com> wrote:
>
> All of the existing callers of remove_dts_thermal_zone() pass a valid
> pointer as the argument, so checking for the NULL pointer is redundant.

True.

> Plus, when calling remove_dts_thermal_zone() from
> intel_soc_dts_iosf_init(), it is possible that
> 1. dts->tzone is an error pointer, when the sensor fails to be
>    registered as a valid thermal zone
> 2. dts->tzone is unregistered in add_dts_thermal_zone(), when some
>    failure occurs after thermal zone registered
> In both cases, there is no need to unregister dts->tzone in
> remove_dts_thermal_zone().
>
> Clear dst->tzone when add_dts_thermal_zone() fails. And do thermal zone
> removal in remove_dts_thermal_zone() only when dts->tzone is set.

Well, I'm not sure.

thermal_zone_device_unregister() will do nothing if the thermal zone
is not really registered AFAICS and it is prudent to restore
SOC_DTS_OFFSET_ENABLE on failure IMO.


> Signed-off-by: Zhang Rui <rui.zhang@intel.com>
> ---
>  drivers/thermal/intel/intel_soc_dts_iosf.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/thermal/intel/intel_soc_dts_iosf.c b/drivers/thermal/intel/intel_soc_dts_iosf.c
> index 7a66d0f077b0..c5203ba8f0b9 100644
> --- a/drivers/thermal/intel/intel_soc_dts_iosf.c
> +++ b/drivers/thermal/intel/intel_soc_dts_iosf.c
> @@ -212,7 +212,7 @@ static int soc_dts_enable(int id)
>
>  static void remove_dts_thermal_zone(struct intel_soc_dts_sensor_entry *dts)
>  {
> -       if (dts) {
> +       if (dts->tzone) {
>                 iosf_mbi_write(BT_MBI_UNIT_PMC, MBI_REG_WRITE,
>                                SOC_DTS_OFFSET_ENABLE, dts->store_status);
>                 thermal_zone_device_unregister(dts->tzone);
> @@ -277,6 +277,7 @@ static int add_dts_thermal_zone(int id, struct intel_soc_dts_sensor_entry *dts,
>  err_enable:
>         thermal_zone_device_unregister(dts->tzone);
>  err_ret:
> +       dts->tzone = NULL;
>         return ret;
>  }
>
> --
> 2.34.1
>

Re: [PATCH] thermal: intel: intel_soc_dts_iosf: Fix thermal_zone removal

[Zhang, Rui]

On Fri, 2023-08-11 at 18:55 +0200, Rafael J. Wysocki wrote:
> On Fri, Aug 11, 2023 at 10:45 AM Zhang Rui <rui.zhang@intel.com>
> wrote:
> > 
> > All of the existing callers of remove_dts_thermal_zone() pass a
> > valid
> > pointer as the argument, so checking for the NULL pointer is
> > redundant.
> 
> True.
> 
> > Plus, when calling remove_dts_thermal_zone() from
> > intel_soc_dts_iosf_init(), it is possible that
> > 1. dts->tzone is an error pointer, when the sensor fails to be
> >    registered as a valid thermal zone
> > 2. dts->tzone is unregistered in add_dts_thermal_zone(), when some
> >    failure occurs after thermal zone registered
> > In both cases, there is no need to unregister dts->tzone in
> > remove_dts_thermal_zone().
> > 
> > Clear dst->tzone when add_dts_thermal_zone() fails. And do thermal
> > zone
> > removal in remove_dts_thermal_zone() only when dts->tzone is set.
> 
> Well, I'm not sure.
> 
> thermal_zone_device_unregister() will do nothing if the thermal zone
> is not really registered AFAICS

yes.

>  and it is prudent to restore
> SOC_DTS_OFFSET_ENABLE on failure IMO.

agreed. Refreshed patch sent.

As this becomes really trivial, feel free to apply it or drop it.

thanks,
rui