From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jon Hunter Subject: Re: [PATCH V3 1/2] Documentation: DT: Add binding documentation for NVIDIA ADMA Date: Tue, 20 Oct 2015 09:54:21 +0100 Message-ID: <5626013D.80602@nvidia.com> References: <1444980919-20331-1-git-send-email-jonathanh@nvidia.com> <1444980919-20331-2-git-send-email-jonathanh@nvidia.com> <56212139.4080101@wwwdotorg.org> <5624D28B.70803@nvidia.com> <56251B56.70007@wwwdotorg.org> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <56251B56.70007-3lzwWm7+Weoh9ZMKESR00Q@public.gmane.org> Sender: linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Stephen Warren Cc: Laxman Dewangan , Vinod Koul , Thierry Reding , Alexandre Courbot , Rob Herring , Pawel Moll , Mark Rutland , Ian Campbell , Kumar Gala , dmaengine-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-tegra@vger.kernel.org On 19/10/15 17:33, Stephen Warren wrote: > On 10/19/2015 05:22 AM, Jon Hunter wrote: >> >> On 16/10/15 17:09, Stephen Warren wrote: >>> On 10/16/2015 01:35 AM, Jon Hunter wrote: >>>> Add device-tree binding documentation for the Tegra210 Audio DMA >>>> controller. >>> >>>> diff --git a/Documentation/devicetree/bindings/dma/tegra210-adma.txt >>>> b/Documentation/devicetree/bindings/dma/tegra210-adma.txt >>> >>>> +Required properties: >>> >>>> +- interrupt-parent: Phandle to the interrupt parent controller. >>> >>> Nit: Since that is more of a "system level"/standard property, it's >>> typical not to document it. The property is not actually required if the >>> inherited value is already correct. Still, it's obvious enough what this >>> means, so I'd only suggest fixing this if you have to respin for some >>> other reason. >> >> Ok. >> >>>> +- clocks: Must contain one entry for the ADMA module clock, >>>> "adma_ape". >>>> +- clock-names: Must contain the entry "adma_ape". >>> >>> Which clock is this in the CAR? I don't see any adma_ape clock >>> documented in the TRM. >> >> Darn. I thought I had checked this. Yes it should be the AHUB clock and >> looking at the current t210 clock patches for mainline this is >> TEGRA210_CLK_D_AUDIO. Ok will fix this. >> >>> Is there a dedicated reset signal for this module? If so, we should >>> require a resets property. >> >> No there does not appear to be. Looking at the documentation the ADMA >> would be reset by the main APE reset. Seems to be one reset that resets >> most modules in the APE. >> >> That raises another issue, the ADMA is in the audio power partition and >> currently our downstream driver assumes that this is on. I should at >> least check this. > > IIRC the situation downstream w.r.t. the audio power partition is > something like: > > The AGIC is in that power partition. The Linux kernel AGIC driver gets > probed before the power domain driver for that domain. (I think) > deferred probe doesn't work for the AGIC driver for some reason (perhaps > this is just a bug or oversight in our downstream AGIC driver?). So, the > AGIC driver can't turn on the audio power domain. So, there's a hack in > the bootloader to turn the power domain on. So, everything in the kernel > assumes that power domain is on at boot and so drivers don't have to > explicitly control/request that power domain. > > We should fix this properly so that upstream doesn't make the assumption > that the audio power domain is magically on before the kernel boots. Yes, I agree. Ok, I will look at that some more. Jon