Currently the code checks the interval value when the temperature is
read which is bad for two reasons:
- checking and setting the interval in the get_temp callback is
inaccurate and awful, that can be done when changing the value.
- Changing the thermal zone structure internals is an abuse of the
exported structure, moreover no lock is taken here.
The goal of this patch is to solve the first item by using the 'set'
function called when changing the interval. The check is done there
and removed from the get_temp function. If the thermal zone was not
initialized yet, the interval is not updated in this case as that will
happen in the init function when registering the thermal zone device.
I don't have any hardware to test the changes.
Signed-off-by: Daniel Lezcano <[email protected]>
---
V2:
- Fixed static function annotation
---
drivers/platform/x86/acerhdf.c | 22 ++++++++++++++++++----
1 file changed, 18 insertions(+), 4 deletions(-)
diff --git a/drivers/platform/x86/acerhdf.c b/drivers/platform/x86/acerhdf.c
index 19fc8ff2225c..b6aa6e5514f4 100644
--- a/drivers/platform/x86/acerhdf.c
+++ b/drivers/platform/x86/acerhdf.c
@@ -334,7 +334,10 @@ static void acerhdf_check_param(struct thermal_zone_device *thermal)
}
if (verbose)
pr_notice("interval changed to: %d\n", interval);
- thermal->polling_delay = interval*1000;
+
+ if (thermal)
+ thermal->polling_delay = interval*1000;
+
prev_interval = interval;
}
}
@@ -349,8 +352,6 @@ static int acerhdf_get_ec_temp(struct thermal_zone_device *thermal, int *t)
{
int temp, err = 0;
- acerhdf_check_param(thermal);
-
err = acerhdf_get_temp(&temp);
if (err)
return err;
@@ -823,8 +824,21 @@ MODULE_ALIAS("dmi:*:*Acer*:pnExtensa*5420*:");
module_init(acerhdf_init);
module_exit(acerhdf_exit);
+static int interval_set_uint(const char *val, const struct kernel_param *kp)
+{
+ int ret;
+
+ ret = param_set_uint(val, kp);
+ if (ret)
+ return ret;
+
+ acerhdf_check_param(thz_dev);
+
+ return 0;
+}
+
static const struct kernel_param_ops interval_ops = {
- .set = param_set_uint,
+ .set = interval_set_uint,
.get = param_get_uint,
};
--
2.25.1
3. Dezember 2020 08:17, "Daniel Lezcano" <[email protected]> schrieb:
> Currently the code checks the interval value when the temperature is
> read which is bad for two reasons:
>
> - checking and setting the interval in the get_temp callback is
> inaccurate and awful, that can be done when changing the value.
>
> - Changing the thermal zone structure internals is an abuse of the
> exported structure, moreover no lock is taken here.
>
> The goal of this patch is to solve the first item by using the 'set'
> function called when changing the interval. The check is done there
> and removed from the get_temp function. If the thermal zone was not
> initialized yet, the interval is not updated in this case as that will
> happen in the init function when registering the thermal zone device.
Thanks for your effort. This improves the code, good finding.
> I don't have any hardware to test the changes.
Tests successfully executed on my good old AOA110.
> Signed-off-by: Daniel Lezcano <[email protected]>
Acked-by: Peter Kaestle <[email protected]>
> ---
> V2:
> - Fixed static function annotation
> ---
> drivers/platform/x86/acerhdf.c | 22 ++++++++++++++++++----
> 1 file changed, 18 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/platform/x86/acerhdf.c b/drivers/platform/x86/acerhdf.c
> index 19fc8ff2225c..b6aa6e5514f4 100644
> --- a/drivers/platform/x86/acerhdf.c
> +++ b/drivers/platform/x86/acerhdf.c
> @@ -334,7 +334,10 @@ static void acerhdf_check_param(struct thermal_zone_device *thermal)
> }
> if (verbose)
> pr_notice("interval changed to: %d\n", interval);
> - thermal->polling_delay = interval*1000;
> +
> + if (thermal)
> + thermal->polling_delay = interval*1000;
> +
> prev_interval = interval;
> }
> }
> @@ -349,8 +352,6 @@ static int acerhdf_get_ec_temp(struct thermal_zone_device *thermal, int *t)
> {
> int temp, err = 0;
>
> - acerhdf_check_param(thermal);
> -
> err = acerhdf_get_temp(&temp);
> if (err)
> return err;
> @@ -823,8 +824,21 @@ MODULE_ALIAS("dmi:*:*Acer*:pnExtensa*5420*:");
> module_init(acerhdf_init);
> module_exit(acerhdf_exit);
>
> +static int interval_set_uint(const char *val, const struct kernel_param *kp)
> +{
> + int ret;
> +
> + ret = param_set_uint(val, kp);
> + if (ret)
> + return ret;
> +
> + acerhdf_check_param(thz_dev);
> +
> + return 0;
> +}
> +
> static const struct kernel_param_ops interval_ops = {
> - .set = param_set_uint,
> + .set = interval_set_uint,
> .get = param_get_uint,
> };
>
> --
> 2.25.1
On 03/12/2020 22:22, Peter Kästle wrote:
> 3. Dezember 2020 08:17, "Daniel Lezcano" <[email protected]> schrieb:
>
>> Currently the code checks the interval value when the temperature is
>> read which is bad for two reasons:
>>
>> - checking and setting the interval in the get_temp callback is
>> inaccurate and awful, that can be done when changing the value.
>>
>> - Changing the thermal zone structure internals is an abuse of the
>> exported structure, moreover no lock is taken here.
>>
>> The goal of this patch is to solve the first item by using the 'set'
>> function called when changing the interval. The check is done there
>> and removed from the get_temp function. If the thermal zone was not
>> initialized yet, the interval is not updated in this case as that will
>> happen in the init function when registering the thermal zone device.
>
> Thanks for your effort. This improves the code, good finding.
>
>
>> I don't have any hardware to test the changes.
>
> Tests successfully executed on my good old AOA110.
>
>
>> Signed-off-by: Daniel Lezcano <[email protected]>
>
> Acked-by: Peter Kaestle <[email protected]>
Thanks for testing the changes.
Shall pick the patches through the thermal tree ?
--
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs
Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
Hi,
On 12/4/20 12:43 PM, Daniel Lezcano wrote:
> On 03/12/2020 22:22, Peter Kästle wrote:
>> 3. Dezember 2020 08:17, "Daniel Lezcano" <[email protected]> schrieb:
>>
>>> Currently the code checks the interval value when the temperature is
>>> read which is bad for two reasons:
>>>
>>> - checking and setting the interval in the get_temp callback is
>>> inaccurate and awful, that can be done when changing the value.
>>>
>>> - Changing the thermal zone structure internals is an abuse of the
>>> exported structure, moreover no lock is taken here.
>>>
>>> The goal of this patch is to solve the first item by using the 'set'
>>> function called when changing the interval. The check is done there
>>> and removed from the get_temp function. If the thermal zone was not
>>> initialized yet, the interval is not updated in this case as that will
>>> happen in the init function when registering the thermal zone device.
>>
>> Thanks for your effort. This improves the code, good finding.
>>
>>
>>> I don't have any hardware to test the changes.
>>
>> Tests successfully executed on my good old AOA110.
>>
>>
>>> Signed-off-by: Daniel Lezcano <[email protected]>
>>
>> Acked-by: Peter Kaestle <[email protected]>
>
> Thanks for testing the changes.
>
> Shall pick the patches through the thermal tree ?
I can take them through the drivers/platform/x86 (pdx86) tree,
but if you prefer to take them upstream through the thermal tree,
then that is fine too...
Here is my ack (as pdx86 maintainer) for taking them through
the thermal tree:
Acked-by: Hans de Goede <[email protected]>
Regards,
Hans
On 07/12/2020 15:54, Hans de Goede wrote:
> Hi,
>
> On 12/4/20 12:43 PM, Daniel Lezcano wrote:
>> On 03/12/2020 22:22, Peter Kästle wrote:
>>> 3. Dezember 2020 08:17, "Daniel Lezcano" <[email protected]> schrieb:
>>>
>>>> Currently the code checks the interval value when the temperature is
>>>> read which is bad for two reasons:
>>>>
>>>> - checking and setting the interval in the get_temp callback is
>>>> inaccurate and awful, that can be done when changing the value.
>>>>
>>>> - Changing the thermal zone structure internals is an abuse of the
>>>> exported structure, moreover no lock is taken here.
>>>>
>>>> The goal of this patch is to solve the first item by using the 'set'
>>>> function called when changing the interval. The check is done there
>>>> and removed from the get_temp function. If the thermal zone was not
>>>> initialized yet, the interval is not updated in this case as that will
>>>> happen in the init function when registering the thermal zone device.
>>>
>>> Thanks for your effort. This improves the code, good finding.
>>>
>>>
>>>> I don't have any hardware to test the changes.
>>>
>>> Tests successfully executed on my good old AOA110.
>>>
>>>
>>>> Signed-off-by: Daniel Lezcano <[email protected]>
>>>
>>> Acked-by: Peter Kaestle <[email protected]>
>>
>> Thanks for testing the changes.
>>
>> Shall pick the patches through the thermal tree ?
>
> I can take them through the drivers/platform/x86 (pdx86) tree,
> but if you prefer to take them upstream through the thermal tree,
> then that is fine too...
>
> Here is my ack (as pdx86 maintainer) for taking them through
> the thermal tree:
>
> Acked-by: Hans de Goede <[email protected]>
Thanks. I'll take them through the thermal tree.
-- Daniel
--
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs
Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog