Re: [PATCH] mfd: tps65219: Add support for soft shutdown via sys-off API

[jerome Neanne <jneanne@baylibre.com>]

On 01/03/2023 17:35, Andrew Davis wrote:
> On 3/1/23 10:07 AM, Lee Jones wrote:
>> On Fri, 03 Feb 2023, Jerome Neanne wrote:
>>
>>> Use new API for power-off mode support:
>>> Link: https://lwn.net/Articles/894511/
>>> Link: https://lore.kernel.org/all/7hfseqa7l0.fsf@baylibre.com/
>>>
>>> sys-off API allows support of shutdown handler and restart handler.
>>>
>>> Shutdown was not supported before that enhancement.
>>> This is required for platform that are not using PSCI.
>>>
> 
> Not sure what platform doesn't have PSCI off, since you tested on
> AM62-SK I'm guessing you manually disabled the PSCI off for testing?
> 
> Anyway I don't see any huge issues with the code itself, small comment 
> below.
> 
>>> Test:
>>> - restart:
>>>    # reboot
>>>    Default is cold reset:
>>>    # cat /sys/kernel/reboot/mode
>>>    Switch boot mode to warm reset:
>>>    # echo warm > /sys/kernel/reboot/mode
>>> - power-off:
>>>    # halt
>>>
>>> Tested on AM62-SP-SK board.
>>>
>>> Signed-off-by: Jerome Neanne <jneanne@baylibre.com>
>>> Suggested-by: Andrew Davis <afd@ti.com>
>>
>> A review from Andrew would be helpful here.
>>
>>> ---
>>>   drivers/mfd/tps65219.c | 45 +++++++++++++++++++++++++++++++-----------
>>>   1 file changed, 34 insertions(+), 11 deletions(-)
>>>
>>> diff --git a/drivers/mfd/tps65219.c b/drivers/mfd/tps65219.c
>>> index 0e402fda206b..c134f3f6e202 100644
>>> --- a/drivers/mfd/tps65219.c
>>> +++ b/drivers/mfd/tps65219.c
>>> @@ -25,25 +25,34 @@ static int tps65219_cold_reset(struct tps65219 *tps)
>>>                     TPS65219_MFP_COLD_RESET_I2C_CTRL_MASK);
>>>   }
>>> -static int tps65219_restart(struct notifier_block *this,
>>> -                unsigned long reboot_mode, void *cmd)
>>> +static int tps65219_soft_shutdown(struct tps65219 *tps)
>>>   {
>>> -    struct tps65219 *tps;
>>> +    return regmap_update_bits(tps->regmap, TPS65219_REG_MFP_CTRL,
>>> +                  TPS65219_MFP_I2C_OFF_REQ_MASK,
>>> +                  TPS65219_MFP_I2C_OFF_REQ_MASK);
>>> +}
>>> -    tps = container_of(this, struct tps65219, nb);
>>> +static int tps65219_power_off_handler(struct sys_off_data *data)
>>> +{
>>> +    tps65219_soft_shutdown(data->cb_data);
>>> +    return NOTIFY_DONE;
>>> +}
>>> +static int tps65219_restart(struct tps65219 *tps,
>>> +                unsigned long reboot_mode)
>>> +{
>>>       if (reboot_mode == REBOOT_WARM)
>>>           tps65219_warm_reset(tps);
>>>       else
>>>           tps65219_cold_reset(tps);
>>> -
>>>       return NOTIFY_DONE;
>>>   }
>>> -static struct notifier_block pmic_rst_restart_nb = {
>>> -    .notifier_call = tps65219_restart,
>>> -    .priority = 200,
>>> -};
>>> +static int tps65219_restart_handler(struct sys_off_data *data)
>>> +{
>>> +    tps65219_restart(data->cb_data, data->mode);
>>> +    return NOTIFY_DONE;
>>> +}
>>>   static const struct resource tps65219_pwrbutton_resources[] = {
>>>       DEFINE_RES_IRQ_NAMED(TPS65219_INT_PB_FALLING_EDGE_DETECT, 
>>> "falling"),
>>> @@ -269,13 +278,27 @@ static int tps65219_probe(struct i2c_client 
>>> *client)
>>>           }
>>>       }
>>> -    tps->nb = pmic_rst_restart_nb;
>>> -    ret = register_restart_handler(&tps->nb);
>>> +    ret = devm_register_sys_off_handler(tps->dev,
>>> +                        SYS_OFF_MODE_RESTART,
>>> +                        SYS_OFF_PRIO_HIGH,
> 
> Why not default prio? SYS_OFF_PRIO_DEFAULT
I'm not completely clear about PRIO recommendations. Will follow your 
suggestion.
> 
> Then you can use this new helper devm_register_restart_handler()
Sure!
> 
>>> +                        tps65219_restart_handler,
>>> +                        tps);
>>> +
>>>       if (ret) {
>>>           dev_err(tps->dev, "cannot register restart handler, %d\n", 
>>> ret);
>>>           return ret;
>>>       }
>>> +    ret = devm_register_sys_off_handler(tps->dev,
>>> +                        SYS_OFF_MODE_POWER_OFF,
>>> +                        SYS_OFF_PRIO_DEFAULT,
>>> +                        tps65219_power_off_handler,
>>> +                        tps);
> 
> 
> devm_register_power_off_handler()?
> 
Oh yes, right, this is solving the PRIO question by construction. This 
is definitely a better option
> Otherwise I see no major issues,
> 
> Reviewed-by: Andrew Davis <afd@ti.com>
> 
> Andrew
> 
>>> +    if (ret) {
>>> +        dev_err(tps->dev, "failed to register sys-off handler: %d\n",
>>> +            ret);
>>> +        return ret;
>>> +    }
>>>       return 0;
>>>   }
>>> -- 
>>> 2.34.1
>>>
>>