Subject: Re: LEDs that change brightness "itself" -- that's a trigger. Re: PM
 regression with LED changes in next-20161109
To: Hans de Goede <hdegoede@redhat.com>, Pavel Machek <pavel@ucw.cz>
Cc: Jacek Anaszewski <jacek.anaszewski@gmail.com>,
        Tony Lindgren <tony@atomide.com>, linux-leds@vger.kernel.org,
        linux-omap@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        linux-kernel@vger.kernel.org, Darren Hart <dvhart@infradead.org>
From: Jacek Anaszewski <j.anaszewski@samsung.com>
Message-id: <9e7d5e0f-c4ca-6930-63b9-83dc28517f33@samsung.com>
Date: Tue, 15 Nov 2016 14:28:39 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
        Thunderbird/45.2.0
MIME-version: 1.0
In-reply-to: <aeccbb02-6e33-390c-d72d-2a580dd67ed8@redhat.com>
Content-type: text/plain; charset=windows-1252; format=flowed
Content-transfer-encoding: 7bit
CMS-TYPE: 201P
References: <cb0aa3ce-ee48-66b8-510c-bd70917afec9@redhat.com>
 <b5ae20d7-8394-7ad5-1830-13173e9105c4@gmail.com>
 <9b476f85-d45e-deb6-335d-fc56f6d90350@redhat.com>
 <127cdd42-6fd8-c671-60b7-3826b351577f@samsung.com>
 <15cafbf5-d842-e184-2fd4-65f8272f505a@redhat.com>
 <CGME20161115103141epcas2p251f2521354a8cd7b9dfcfe55bfa482f0@epcas2p2.samsung.com>
 <20161115103133.GA22860@amd>
 <4e392d5d-eb10-f285-517e-976a55c3e318@samsung.com> <20161115111154.GA5482@amd>
 <128aae59-b790-42f1-7d66-81391c9330c3@redhat.com> <20161115114859.GA7018@amd>
 <aeccbb02-6e33-390c-d72d-2a580dd67ed8@redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2997
Lines: 83

On 11/15/2016 01:06 PM, Hans de Goede wrote:
> Hi,
>
> On 15-11-16 12:48, Pavel Machek wrote:
>> Hi!
>>
>>>>>> The LED you are talking about _has_ a trigger, implemented in
>>>>>> hardware. That trigger can change LED brightness behind kernel's (and
>>>>>> userspace's) back. Don't pretend the trigger does not exist, it does.
>>>>>>
>>>>>> And when you do that, you'll have nice place to report changes to
>>>>>> userspace -- trigger can now export that information, and offer
>>>>>> poll()
>>>>>> interface.
>>>>>
>>>>> Well, that sounds interesting. It is logically justifiable.
>>>>
>>>> Thanks.
>>>>
>>>>> I initially proposed exactly this solution, with recently
>>>>> added userspace LED being a trigger listener. It seems a bit
>>>>> awkward though. How would you listen to the trigger events?
>>>>
>>>> Trigger exposes a file in sysfs, with poll() working on that file
>>>
>>> Hmm, a new file would give the advantage of making it easy for
>>> userspace to see if the trigger is poll-able, this is likely
>>> better then my own proposal I just send.
>>
>> Good.
>>
>>>> (and
>>>> probably read exposing the current brightness).
>>>
>>> If we do this, can we please make it mirror brightness, iow
>>> also make it writable, that will make it easier for userspace
>>> to deal with it. We can simply re-use the existing show / store
>>> methods for brightness for this.
>>
>> Actually, echo 0 > brightness disables the trigger, IIRC. I'd avoid
>> that here, you want to be able to turn off the backlight but still
>> keep the trigger (and be notified of future changes).
>
> True, that is easy to do the store method will just need to call
> led_set_brightness_nosleep instead of led_set_brightness, this
> will skip the checks to stop blinking in led_set_brightness and
> otherwise is equivalent.
>
>>> I suggest we call it:
>>>
>>> trigger_brightness
>>>
>>> And only register it when a poll-able trigger is present.
>>
>> I'd call it 'current_brightness', but that's no big deal. Yes, only
>> registering it for poll-able triggers makes sense.
>
> current_brightness works for me. I will take a shot a patch-set
> implementing this.

Word "current" is not precise here.

It can be thought of as either last brightness set by the
user or the brightness currently written to the device
(returned by brightness file).

There is a semantic discrepancy in our requirements -
we want the file representing both permanent brightness
set by the user and brightness set by the hardware.

The two stand in contradiction to each other since
brightness set by the user can be adjusted by the hardware.

Reading the file shouldn't update brightness property of
struct led_classdev, so it shouldn't call led_update_brightness()
but it still should allow reading brightness set by the
hardware, as a result of each POLLPRI event. So in fact in
the same time it should report both according to our requirements
which is impossible. Do we need three brightness files ?

-- 
Best regards,
Jacek Anaszewski