From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Message-ID: <4F8EB784.6090702@kernel.org> Date: Wed, 18 Apr 2012 13:45:56 +0100 From: Jonathan Cameron MIME-Version: 1.0 To: "Getz, Robin" CC: Lars-Peter Clausen , Jonathan Cameron , "linux-iio@vger.kernel.org" , Marten Svanfeldt Subject: Re: [PATCH v3] staging:iio: trigger: Add hrtimer trigger References: <1334581381-18161-1-git-send-email-lars@metafoo.de> <4a3800e1-986a-4801-87a4-f0246425c87f@email.android.com> <4F8C4F17.5040101@metafoo.de> <201204180758.34751.robin.getz@analog.com> In-Reply-To: <201204180758.34751.robin.getz@analog.com> Content-Type: text/plain; charset=UTF-8; format=flowed List-ID: On 4/18/2012 12:58 PM, Getz, Robin wrote: > On Mon 16 Apr 2012 12:55, Lars-Peter Clausen pondered: >> On 04/16/2012 06:17 PM, Jonathan Cameron wrote: >>> Lars-Peter Clausen wrote: >>>> From: Marten Svanfeldt >>>> >>>> This patch adds a IIO trigger driver which uses a highres timer to >>>> provide a >>>> frequency based trigger. >>> Fine as it stands but same issue arises as we had with userspace trigger >>> still. What are we doing registering a pure software element not >>> associated to any specific hardware via a platform device. Why not do it >>> on userspace asking for one as we do with the sysfs file based trigger? >> I suppose this is a general question how we want to mange our triggers in >> general. None of the other existing trigger drivers does direct IO access >> and just use existing infrastructure. They could all be easily be >> instantiated by writing a string or number to a sysfs file. So where do we >> draw the line? Fair point. Personally I'd go for whether it is about explicit hardware or not. So gpio / general interrupt makes sense in platform code. I suspect we'll kill off the RTC one anyway on the way out of staging. > Isn't there an issue of accuracy? the timing accuracy of sysfs/userspace is > non-existant with respect to what you need to do in most of these cases. I'm not suggesting there is any problem with having a hrtimer based trigger (in fact I am throughly in favour!) its just a question of whether it should be registered in the board file / device tree or done via a magic string write as Lars-Peter mentions above (which is what we do the sysfs file based trigger precisely because there was a pretty strong feeling against implying non existent hardware...)