From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jacek Anaszewski Subject: Re: [PATCH v2 2/2] leds: Add Spreadtrum SC27xx breathing light controller driver Date: Sat, 12 May 2018 22:44:14 +0200 Message-ID: <7c9d5eec-4e6b-9b26-470c-f8002cca5f27@gmail.com> References: <9a2a07b8eb313ae3ba64af911337ee7ff7c9ad43.1525757122.git.baolin.wang@linaro.org> <20180509142539.GB25131@amd> <20180510113749.GG6977@amd> <6289571e-7224-ca5e-1acf-5b099be57302@gmail.com> <20180512083559.GB8944@amd> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20180512083559.GB8944@amd> Sender: linux-kernel-owner@vger.kernel.org To: Pavel Machek Cc: Baolin Wang , robh+dt@kernel.org, mark.rutland@arm.com, xiaotong.lu@spreadtrum.com, broonie@kernel.org, linux-leds@vger.kernel.org, devicetree@vger.kernel.org, linux-kernel@vger.kernel.org List-Id: devicetree@vger.kernel.org Hi Pavel, On 05/12/2018 10:35 AM, Pavel Machek wrote: > Hi! > >>>> I disagree here. We already had the same discussion at the occasion >>>> of the patch [0] and it turned out to be a dead-end [1]. Now we have >>>> neither the driver nor the generic pattern interface. >>>> >>>> We also already have some older LED class drivers that implement custom >>>> pattern interfaces (e.g. drivers/leds/leds-lm3533.c) and the same >>>> approach can be applied in this case. >>> >>> Please don't. It was mistake to implement custom pattern interfaces >>> back then, it is still mistake now. >> >> It turned out to be really hard to cover all known pattern generator >> implementations with generic interface. Sure, it would be nice to have >> one, but the whole discussion around [0] only unveiled the diversity of >> parameters to cover. And still new devices appear on the market. >> >> We would have to propose a set of pattern schemes and allow to >> add new ones to it. > > I believe that what I'm proposing below is close enough to universal. > >>> If we really need solution now, I'd recommend "pattern" file with >>> >>> " ". >>> >>> In this specific case, hardware only supports patterns in this format: >>> >>> low_time 0 rise_time 255 high_time 255 fall_time 0 >>> >>> so driver would simply -EINVAL on anything else. >> >> I'm fine with the pattern file, but the pattern format would have >> to be defined in the per-driver ABI documentation. It wouldn't much >> differ from the custom pattern approach though, unless I'm missing some >> gain of having pattern setting in a uniformly named single sysfs file >> (with semantics differing from driver to driver). > > I'm proposing " ..." sysfs file. It certainly > covers this hardware, it would be enough to cover the Qualcomm Pulse > generator (IIRC), and it would cover most uses cases of Nokia N900's > LED. > > Yes, we would need to document limitations of each chip. But it should > be easily possible to run pattern designed for Spreadtrum on N900, > even if it would not work the other way around. > > (If someone really wants to run complex patterns on simple hardware, > we can provide software emulation using same file format. I believe I > still have that patch somewhere.) OK, I've revised the discussion under Qualcomm LPG patch set and it seems that we have almost ready solution in [0], except the pattern_repeat file you mention in [1]. So probably Baolin could address your remarks from [1] and add pattern_repeat file to the patch that begins thread [0]. [0] https://lkml.org/lkml/2017/11/15/27 [1] https://lkml.org/lkml/2017/12/8/470 -- Best regards, Jacek Anaszewski