Re: [PATCH 1/2] hwmon: pwm-fan: Add profile support and add remove module support

[Guenter Roeck]

On 5/26/20 2:05 AM, Sandipan Patra wrote:
> Hi Uwe,
> 
> 
> 
>> -----Original Message-----
>> From: Uwe Kleine-KÃnig <u.kleine-koenig@xxxxxxxxxxxxxx>
>> Sent: Tuesday, May 26, 2020 12:35 PM
>> To: Sandipan Patra <spatra@xxxxxxxxxx>
>> Cc: Thierry Reding <treding@xxxxxxxxxx>; Jonathan Hunter
>> <jonathanh@xxxxxxxxxx>; kamil@xxxxxxxxx; jdelvare@xxxxxxxx; linux@roeck-
>> us.net; robh+dt@xxxxxxxxxx; Bibek Basu <bbasu@xxxxxxxxxx>; Bitan Biswas
>> <bbiswas@xxxxxxxxxx>; linux-pwm@xxxxxxxxxxxxxxx; linux-
>> hwmon@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-
>> tegra@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
>> Subject: Re: [PATCH 1/2] hwmon: pwm-fan: Add profile support and add remove
>> module support
>>
>> External email: Use caution opening links or attachments
>>
>>
>> On Tue, May 26, 2020 at 10:36:04AM +0530, Sandipan Patra wrote:
>>> This change has 2 parts:
>>> 1. Add support for profiles mode settings.
>>>     This allows different fan settings for trip point temp/hyst/pwm.
>>>     T194 has multiple fan-profiles support.
>>>
>>> 2. Add pwm-fan remove support. This is essential since the config is
>>>     tristate capable.
>>
>> These two are orthogonal, aren't they? So they belong in two patches.
>>
>> You have to expand the binding documentation.
>>
>>> Signed-off-by: Sandipan Patra <spatra@xxxxxxxxxx>
>>> ---
>>>  drivers/hwmon/pwm-fan.c | 112
>>> ++++++++++++++++++++++++++++++++++++++++++------
>>>  1 file changed, 100 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/drivers/hwmon/pwm-fan.c b/drivers/hwmon/pwm-fan.c index
>>> 30b7b3e..26db589 100644
>>> --- a/drivers/hwmon/pwm-fan.c
>>> +++ b/drivers/hwmon/pwm-fan.c
>>> @@ -3,8 +3,10 @@
>>>   * pwm-fan.c - Hwmon driver for fans connected to PWM lines.
>>>   *
>>>   * Copyright (c) 2014 Samsung Electronics Co., Ltd.
>>> + * Copyright (c) 2020, NVIDIA Corporation.
>>>   *
>>>   * Author: Kamil Debski <k.debski@xxxxxxxxxxx>
>>> + * Author: Sandipan Patra <spatra@xxxxxxxxxx>
>>>   */
>>>
>>>  #include <linux/hwmon.h>
>>> @@ -21,6 +23,8 @@
>>>  #include <linux/timer.h>
>>>
>>>  #define MAX_PWM 255
>>> +/* Based on OF max device tree node name length */
>>> +#define MAX_PROFILE_NAME_LENGTH      31
>>>
>>>  struct pwm_fan_ctx {
>>>       struct mutex lock;
>>> @@ -38,6 +42,12 @@ struct pwm_fan_ctx {
>>>       unsigned int pwm_fan_state;
>>>       unsigned int pwm_fan_max_state;
>>>       unsigned int *pwm_fan_cooling_levels;
>>> +
>>> +     unsigned int pwm_fan_profiles;
>>> +     const char **fan_profile_names;
>>> +     unsigned int **fan_profile_cooling_levels;
>>> +     unsigned int fan_current_profile;
>>> +
>>>       struct thermal_cooling_device *cdev;  };
>>>
>>> @@ -227,28 +237,86 @@ static int pwm_fan_of_get_cooling_data(struct
>> device *dev,
>>>                                      struct pwm_fan_ctx *ctx)  {
>>>       struct device_node *np = dev->of_node;
>>> +     struct device_node *base_profile = NULL;
>>> +     struct device_node *profile_np = NULL;
>>> +     const char *default_profile = NULL;
>>>       int num, i, ret;
>>>
>>> -     if (!of_find_property(np, "cooling-levels", NULL))
>>> -             return 0;
>>> +     num = of_property_count_u32_elems(np, "cooling-levels");
>>> +     if (num <= 0) {
>>> +             base_profile = of_get_child_by_name(np, "profiles");
>>> +             if (!base_profile) {
>>> +                     dev_err(dev, "Wrong Data\n");
>>> +                     return -EINVAL;
>>> +             }
>>> +     }
>>> +
>>> +     if (base_profile) {
>>> +             ctx->pwm_fan_profiles =
>>> +                     of_get_available_child_count(base_profile);
>>> +
>>> +             if (ctx->pwm_fan_profiles <= 0) {
>>> +                     dev_err(dev, "Profiles used but not defined\n");
>>> +                     return -EINVAL;
>>> +             }
>>>
>>> -     ret = of_property_count_u32_elems(np, "cooling-levels");
>>> -     if (ret <= 0) {
>>> -             dev_err(dev, "Wrong data!\n");
>>> -             return ret ? : -EINVAL;
>>> +             ctx->fan_profile_names = devm_kzalloc(dev,
>>> +                     sizeof(const char *) * ctx->pwm_fan_profiles,
>>> +                                                     GFP_KERNEL);
>>> +             ctx->fan_profile_cooling_levels = devm_kzalloc(dev,
>>> +                     sizeof(int *) * ctx->pwm_fan_profiles,
>>> +                                                     GFP_KERNEL);
>>> +
>>> +             if (!ctx->fan_profile_names
>>> +                             || !ctx->fan_profile_cooling_levels)
>>> +                     return -ENOMEM;
>>> +
>>> +             ctx->fan_current_profile = 0;
>>> +             i = 0;
>>> +             for_each_available_child_of_node(base_profile, profile_np) {
>>> +                     num = of_property_count_u32_elems(profile_np,
>>> +                                                     "cooling-levels");
>>> +                     if (num <= 0) {
>>> +                             dev_err(dev, "No data in cooling-levels inside profile
>> node!\n");
>>> +                             return -EINVAL;
>>> +                     }
>>> +
>>> +                     of_property_read_string(profile_np, "name",
>>> +                                             &ctx->fan_profile_names[i]);
>>> +                     if (default_profile &&
>>> +                             !strncmp(default_profile,
>>> +                             ctx->fan_profile_names[i],
>>> +                             MAX_PROFILE_NAME_LENGTH))
>>> +                             ctx->fan_current_profile = i;
>>> +
>>> +                     ctx->fan_profile_cooling_levels[i] =
>>> +                             devm_kzalloc(dev, sizeof(int) * num,
>>> +                                                     GFP_KERNEL);
>>> +                     if (!ctx->fan_profile_cooling_levels[i])
>>> +                             return -ENOMEM;
>>> +
>>> +                     of_property_read_u32_array(profile_np, "cooling-levels",
>>> +                             ctx->fan_profile_cooling_levels[i], num);
>>> +                     i++;
>>> +             }
>>>       }
>>>
>>> -     num = ret;
>>>       ctx->pwm_fan_cooling_levels = devm_kcalloc(dev, num, sizeof(u32),
>>>                                                  GFP_KERNEL);
>>>       if (!ctx->pwm_fan_cooling_levels)
>>>               return -ENOMEM;
>>>
>>> -     ret = of_property_read_u32_array(np, "cooling-levels",
>>> -                                      ctx->pwm_fan_cooling_levels, num);
>>> -     if (ret) {
>>> -             dev_err(dev, "Property 'cooling-levels' cannot be read!\n");
>>> -             return ret;
>>> +     if (base_profile) {
>>> +             memcpy(ctx->pwm_fan_cooling_levels,
>>> +               ctx->fan_profile_cooling_levels[ctx->fan_current_profile],
>>> +                                             num);
>>> +     } else {
>>> +             ret = of_property_read_u32_array(np, "cooling-levels",
>>> +                             ctx->pwm_fan_cooling_levels, num);
>>> +             if (ret) {
>>> +                     dev_err(dev, "Property 'cooling-levels' cannot be read!\n");
>>> +                     return -EINVAL;
>>> +             }
>>>       }
>>>
>>>       for (i = 0; i < num; i++) {
>>> @@ -390,6 +458,25 @@ static int pwm_fan_probe(struct platform_device
>> *pdev)
>>>       return 0;
>>>  }
>>>
>>> +static int pwm_fan_remove(struct platform_device *pdev) {
>>> +     struct pwm_fan_ctx *ctx = platform_get_drvdata(pdev);
>>> +     struct pwm_args args;
>>> +
>>> +     if (!ctx)
>>> +             return -EINVAL;
>>> +
>>> +     if (IS_ENABLED(CONFIG_THERMAL))
>>> +             thermal_cooling_device_unregister(ctx->cdev);
>>> +
>>> +     pwm_get_args(ctx->pwm, &args);
>>> +     pwm_config(ctx->pwm, 0, args.period);
>>> +     pwm_disable(ctx->pwm);
>>
>> What is what you really here? Is it only that the PWM stops oscillating, or is it
>> crucial that the output goes to its inactive level?
>>
>> (The intended semantic of pwm_disable includes that the output goes low, but
>> not all implementations enforce this.)
>>
>> Also please don't introduce new users of pwm_config() and pwm_disable() use
>> pwm_apply() instead.
>>
>> I wonder if this unregistration is "safe". When the driver is in use I'd expect that
>> the hwmon device doesn't go away and so the devm unregistration callback that
>> belongs to
>> devm_hwmon_device_register_with_groups() blocks. But at this time the PWM
>> is already stopped and so the device stops functioning earlier.
>>
>> Best regards
>> Uwe
>>
> 
> Thanks for reviewing the changes.
> 
> I see that pwm_fan_shutdown() which has got merged recently, can also be used for
> module remove functionality. May be it will need a little bit of tweak in the code.
> However I should have not made both multiple profiles support and fan remove functionality on
> same patch.
> 

Pointing out explicitly:

ret = devm_add_action_or_reset(dev, pwm_fan_pwm_disable, ctx);

cdev = devm_thermal_of_cooling_device_register(dev, ...)

Guenter