Hi Murali, On 7/28/2012 3:34 AM, Karicheri, Muralidharan wrote: >>> >>> On 7/25/2012 9:16 PM, Karicheri, Muralidharan wrote: >>>> All, >>>> >>>> >>>> >>>> I am working to migrate DaVinci platforms to common clock framework. >>>> While doing so I would need your support in terms of answering any >>>> questions that I have, reviewing patches and testing. I have started >>>> with DM6446 and I got the following questions to you. Could someone >>>> clarify this for me? >>>> >>>> >>>> >>>> 1. davinci_clk_disable_unused(). From the description of the function >>>> header, I understand this was added to disable lpsc modules that are >>>> left enabled by the bootloader. So it appears that this is a hack. >>> >>> I wouldn't call it a hack. Reducing dependency on bootloader is a >>> desirable feature. >>> >>>> Ideally bootloader needs to disable all of the modules before it enters >>>> the kernel and this function shouldn’t have been there. Can someone >>>> clarify if this is still needed? >>> >>> As you said that's in ideal world. Even if the latest bootloader had the >>> code in place to do this, I suspect many would be stuck with old >>> bootloaders on their boards so this is still required. >>> > > Agreed. But to support this, I think I need to enhance the clk API interface. > There is a function is_enabled() that can be implemented by the driver ( > clk-psc in our case), but how do we know if a clock is in use or not. So > a new API needs to be added is_inuse() or such which returns the enabled > count. Once we have this, we should be able to implement > davinci_clk_disable_unused() > with common clock framework. I will send an RFC for this enhancement if > this is not available already.
I have not looked into the common clock framework so far, so cant comment on the details. OMAP has a function that does similar functionality (see omap2_clk_disable_unused() in arch/arm/mach-omap2/clock.c). Rajendra Nayak has posted some patches[1] converting OMAP over to common clock. May be you could have a look at those to see how he has handled this? > >>>> >>>> 2. In dm644x.c, for dsp_clk, usecount is set to 1 with a comment >>>> “REVISIT how to disable”. This is true for timer2_clk as well as >>>> vicp_clk. I see similar things on other DaVinci platforms. Have someone >>>> investigated why this hack is required? Any volunteers to debug and root >>>> cause this? In common clock framework, we wouldn’t be able to support >>>> this hack and has to be debugged and fixed. Any comments? >>> >>> I have no idea why these were put in. If you drop these hacks on >>> existing kernel, do you see any issues? vicp_clk sounds like a video >>> clock. Cc Manju and Prabhakar for comment on that. > > As I have said before, I need your team's help to test this on DM devices. > I have a DM6446 that I can use to do a boot test. I have most of the code > ready for DM644x and will start testing this next week. Thanks! Please test on whatever platform you have and I will help test on the rest. Once the patches go into linux-next they will get some testing too. Thanks, Sekhar _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
