Re: [RFC PATCH v3 2/2] pwm: imx: Configure output to GPIO in disabled state

["Michal Vokáč" <michal.vokac@ysoft.com>]

On 24.1.2019 10:22, Uwe Kleine-König wrote:
> On Thu, Jan 24, 2019 at 09:59:47AM +0100, Michal Vokáč wrote:
>> On 12.12.2018 13:12, Uwe Kleine-König wrote:
>>> On Wed, Dec 12, 2018 at 11:42:17AM +0000, Vokáč Michal wrote:
>>>> On 12.12.2018 09:01, Uwe Kleine-König wrote:
>>>>> On Thu, Dec 06, 2018 at 01:41:31PM +0000, Vokáč Michal wrote:
>>>>>> Normally the PWM output is held LOW when PWM is disabled. This can cause
>>>>>> problems when inverted PWM signal polarity is needed. With this behavior
>>>>>> the connected circuit is fed by 100% duty cycle instead of being shut-off.
>>>>>>
>>>>>> Allow users to define a "pwm" and a "gpio" pinctrl states. The pwm pinctrl
>>>>>> state is selected when PWM is enabled and the gpio pinctrl state is
>>>>>> selected when PWM is disabled. In the gpio state the new pwm-gpios GPIO is
>>>>>> configured as input and the internal pull-up resistor is used to pull the
>>>>>> output level high.
>>>>>>
>>>>>> If all the pinctrl states and the pwm-gpios GPIO are not correctly
>>>>>> specified in DT the PWM work as usual.
>>>>>>
>>>>>> As an example, with this patch a PWM controlled backlight with inversed
>>>>>> signal polarity can be used in full brightness range. Without this patch
>>>>>> the backlight can not be turned off as brightness = 0 disables the PWM
>>>>>> and that in turn set PWM output LOW, that is full brightness.
>>>>>>
>>>>>> Inverted output of the PWM with "default" and with "pwm"+"gpio" pinctrl:
>>>>>>
>>>>>> +--------------+------------+---------------+----------- +-------------+
>>>>>> | After reset  | Bootloader |   PWM probe   |     PWM    |     PWM     |
>>>>>> | 100k pull-up |            |               | enable 30% |   disable   |
>>>>>> +--------------+------------+---------------+------------+-------------+
>>>>>> | pinctrl      |    none    |    default    |   default  |   default   |
>>>>>> | out H   __________________                   __   __                 |
>>>>>> | out L                     \_________________/  \_/  \_/\____________ |
>>>>>> |                                           ^    ^    ^                |
>>>>>> +--------------+------------+---------------+------------+-------------+
>>>>>> | pinctrl      |    none    |      gpio     |     pwm    |     gpio    |
>>>>>> | out H   __________________________________   __   __   _____________ |
>>>>>> | out L                                     \_/  \_/  \_/              |
>>>>>> |                                           ^    ^    ^                |
>>>>>> +----------------------------------------------------------------------+
>>>>>>
>>>>>> Signed-off-by: Michal Vokáč <michal.vokac@ysoft.com>
>>>>>> ---
>>>>>> Changes in v3:
>>>>>>     - Commit message update.
>>>>>>     - Minor fix in code comment (Uwe)
>>>>>>     - Align function arguments to the opening parentheses. (Uwe)
>>>>>>     - Do not test devm_pinctrl_get for NULL. (Thierry)
>>>>>>     - Convert all messages to dev_dbg() (Thierry)
>>>>>>     - Do not actively drive the pin in gpio state. Configure it as input
>>>>>>       and rely solely on the internal pull-up. (Thierry)
>>>>>>
>>>>>> Changes in v2:
>>>>>>     - Utilize the "pwm" and "gpio" pinctrl states.
>>>>>>     - Use the pwm-gpios signal to drive the output in "gpio" pinctrl state.
>>>>>>     - Select the right pinctrl state in probe.
>>>>>>
>>>>>>     drivers/pwm/pwm-imx.c | 77 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>>>     1 file changed, 77 insertions(+)
>>>>>>
>>>>
>>>> [ snip ]
>>>>
>>>>> On thing I noticed while looking at the rcar driver is: This doesn't
>>>>> wait for the current period to end. Is this supposed to happen? Also for
>>>>> the enable case the hardware is configured for the desired duty cycle
>>>>> and only then the pinctrl is switched to pwm. Both might result in a
>>>>> spike that is not desired.
>>>>
>>>> The behavior should not change from how imx-pwm was working before.
>>>> When PWM is disabled the output is immediately gated off (put into
>>>> the idle state) independently on the period. I measured this.
>>>
>>> Oh really, I wasn't aware of that. This is another bug in the imx pwm
>>> then (I think).
>>
>> I kind of expect that when hit a disable button the output is immediately
>> put into the idle state. To me it does not seem appropriate to wait the
>> whole period, or even just the active part of the period. If duty=100%
>> and period=4s (current maximum), in the worst case you would have to wait
>> 4s until you stop the PWM. Quite a long time of driving something you
>> actually wanted to be shut off.
> 
> I think it might be beneficial to allow (or require) that disable acts
> immediately. But this is not how it used to be and in my discussion with
> Thierry (IIRC) he required to complete the currently running period.

I am confused here. IFAIK it always used to be that:

	pwm_apply_state(pwm, { .enabled = 0 });

immediately stops the PWM with:

	writel(0, imx->mmio_base + MX3_PWMCR);

regardless of the period (for pwm-imx).

> Then if a user requires a "smooth disable" (i.e. with completing the
> currently running period) they can do:
> 
> 	pwm_apply_state(pwm, { .duty_cycle = 0, ... });
> 	pwm_apply_state(pwm, { .enabled = 0 });
> 
> to get the delayed disable.
> 
> But having said this, this should be a concious decision with the
> requirements properly documented and the drivers users reviewed and adapted.
> 
>>>> For the enable case you would certainly not get any additional spikes.
>>>
>>> Yes, there is a possibility for a spike: If you configure for say 40%:
>>>                     _    _
>>> pwm output   : ___/ \__/ \__
>>> muxing       : GPIO| PWM_
>>> actual output: ____/\__/ \__
>>
>> OK, you are right.
>>
>>>> The worst thing that may happen is that the first period will be
>>>> slightly shorter depending on how long it takes to test the pinctrl
>>>> and switch the muxing. And this is unavoidable, it would happen even
>>>> if you wait for the start of a period. The test + muxing still takes
>>>> some time.
>>>
>>> You could first configure for duty_cycle 0 and a short period, then mux
>>> to PWM and then configure the correct duty cycle. Ditto for disable.
>>
>> I agree it can be solved for the enable case but I do not see the
>> point in doing so for the disable case. Can you elaborate on how it
>> could be useful?
> 
> Assuming we stick to "disable should complete the currently running
> period" it is all about preventing spikes. So if you switch the mux from
> PWM to GPIO when the output just raised, that might (depending on the
> use case) be bad and so it should be possible for the user to prevent
> this.

I understand. The issue is I do not tend to call this case "a spike".
To me it means that the first and last period may have a shorter duty
cycle than requested. I tend to think that shorter duty cycle is not
harmful/bad. It actually means that you enable/disable the device
more smoothly.

Off course everything depends on the device that is fed with the signal.
So I totaly agree that being able to produce the signal with all the
periods of the same length and with the same duty is a good thing.

>> This is what I came up with:
>>
>> [...]
>>
>> What do you think?
> 
> I'm missing context. Your patch is on top of your last patch I assume.

Sorry, I had to add some context. It is against current linux-next and
it is just the relevant hunk for the pwm_imx27_apply() function.

> Would you mind to send the squashed result?

Of course I do not mind. I can respin and send what I have as v5
against linux-next as I already rebased on top of your patches with
the split driver, if you prefer that?
  
> (I'm unsure when I find time (and nerve) for a proper review, but I will
> try to at least look over your suggestions.)

Thank you very much Uwe!
Michal